spatial dimensions of tower karst and cockpit karst: a

University of Wisconsin MilwaukeeUWM Digital Commons

Theses and Dissertations

December 2014

Spatial Dimensions of Tower Karst and CockpitKarst: A Case Study of Guilin, ChinaWei HuangUniversity of Wisconsin-Milwaukee

Follow this and additional works at: https://dc.uwm.edu/etdPart of the Geographic Information Sciences Commons, and the Geomorphology Commons

This Dissertation is brought to you for free and open access by UWM Digital Commons. It has been accepted for inclusion in Theses and Dissertationsby an authorized administrator of UWM Digital Commons. For more information, please contact [email protected].

Recommended CitationHuang, Wei, "Spatial Dimensions of Tower Karst and Cockpit Karst: A Case Study of Guilin, China" (2014). Theses and Dissertations.626.https://dc.uwm.edu/etd/626

https://dc.uwm.edu/?utm_source=dc.uwm.edu%2Fetd%2F626&utm_medium=PDF&utm_campaign=PDFCoverPages

https://dc.uwm.edu/etd?utm_source=dc.uwm.edu%2Fetd%2F626&utm_medium=PDF&utm_campaign=PDFCoverPages

https://dc.uwm.edu/etd?utm_source=dc.uwm.edu%2Fetd%2F626&utm_medium=PDF&utm_campaign=PDFCoverPages

http://network.bepress.com/hgg/discipline/358?utm_source=dc.uwm.edu%2Fetd%2F626&utm_medium=PDF&utm_campaign=PDFCoverPages

http://network.bepress.com/hgg/discipline/1053?utm_source=dc.uwm.edu%2Fetd%2F626&utm_medium=PDF&utm_campaign=PDFCoverPages

https://dc.uwm.edu/etd/626?utm_source=dc.uwm.edu%2Fetd%2F626&utm_medium=PDF&utm_campaign=PDFCoverPages

mailto:[email protected]

SPATIAL DIMENSIONS OF TOWER KARST AND COCKPIT KARST:

A CASE STUDY OF GUILIN, CHINA

by

Wei Huang

A Dissertation Submitted in

Partial Fulfillment of the

Requirements for the Degree of

Doctor of Philosophy

in Geography

at

The University of Wisconsin-Milwaukee

December 2014

ii

ABSTRACT

SPATIAL DIMENSIONS OF TOWER KARST AND COCKPIT KARST:

A CASE STUDY OF GUILIN, CHINA

by

Wei Huang

The University of Wisconsin-Milwaukee, 2014

Under the Supervision of Professor Michael J. Day

Abstract

Tower karst (fenglin) and cockpit karst (fengcong) are two globally important

representative styles of tropical karst. Previously proposed sequential and parallel

development models are preliminary, and geomorphological studies to date do not

provide enough satisfactory evidence to delineate the spatial and temporal relation

between the two landscapes. This unclear interpretation of tower-cockpit relationships

not only obscures understanding of the process-form dynamics of these tropical karst

landforms, but also confuses their definition. Moreover, previous technological

limitations, as well as the fragmental nature of the karst landscapes, has limited

incorporation of geologic and other data into broad geospatial frameworks based on

geographic information science (GIS) and remote sensing (RS), with such data being

spatially and temporally disparate. This study incorporates various data sources to

iii

address the fenglin-fengcong relationship, particularly the recently postulated “edge

effect”, which has not been examined in detail previously and which may hinge upon the

interaction of multiple environmental variables, including geomorphology, vegetation

and hydrology. To address these issues, this research combines geographic, geologic and

hydrologic data, using GIS and RS technologies to test quantitatively the “edge effect”

hypothesis.

Specifically, there are four inter-related objectives of this study. The first is to develop a

method to effectively differentiate fenglin and fengcong. The second is to extract

optimally the vegetation information from satellite imagery, and investigate the

correlation between tropical karst topography and its vegetation. The third is to combine

the regional hydrologic data and solute transport models to estimate geochemicals control

of fenglin and fengcong. The fourth one, perhaps the most important, is to test the “edge

effect” hypothesis using the results from the other three objectives.

There are several significant conclusions. First, DEM data are very useful for extracting

profiles of complex surface landforms from satellite imagery. Second, the vegetation

distribution varies between tower karst and cockpit karst and the differences correlate

with topographic characteristics. The under-representation of vegetation on the south-

southwest aspect of tower karst is remarkable, and its overall distribution is both less

abundant and dispersed than in cockpit karst. Third, the “edge effect” exists in the Guilin

area, with variable intensity and extension in different dimensions.

In summary, the major contributions of the study include the following. First, the study

has developed a method to classify fenglin-fengcong tropical karst effectively, even with

the presence of shadows that would otherwise hinder traditional classification. Second,

iv

the study showed a variance of vegetation vitality within aspects of fenglin that might

relate to its geomorphic difference from fengcong. Third, the study combined

groundwater and solute transport models to estimate bicarbonate distributions,

representing a novel systematic and quantitative approach to tropical karst studies.

v

© Copyright by Wei Huang, 2014

All Rights Reserved

vi

Dedicated to my parents…

vii

TABLE OF CONTENTS

ABSTRACT .................................................................................................................................... ii

TABLE OF CONTENTS ............................................................................................................ vii

LIST OF FIGURES ...................................................................................................................... ix

LIST OF TABLES ........................................................................................................................ xi

ACKNOWLEDGMENTS ........................................................................................................... xii

CHAPTER 1 INTRODUCTION .................................................................................................. 1

CHAPTER 2 DIFFERENTIATING TOWER KARST (FENGLIN) AND COCKPIT

KARST (FENGCONG) USING DEM CONTOUR, SLOPE AND CENTROID..................... 7

2.1 Introduction .......................................................................................................................... 7

2.2 Study area and data ........................................................................................................... 14

2.3 Methodology ....................................................................................................................... 18

2.4 Results and Discussion ....................................................................................................... 22

2.5 Conclusion .......................................................................................................................... 28

CHAPTER 3 INVESTIGATION OF VEGETATION ON SURFACE KARST USING

TOPOGRAPHIC CORRECTION, SHADOW RETRIEVAL AND VEGETATION INDEX

....................................................................................................................................................... 29

3.1 Introduction ........................................................................................................................ 29


3.3 Methodology ....................................................................................................................... 34


3.5 Conclusion .......................................................................................................................... 44

CHAPTER 4 HYDROLOGIC CONTROLS ON DEVELOPING TOWER KARST AND

COCKPIT KARST ...................................................................................................................... 45

4.1 Introduction ........................................................................................................................ 45


4.3 Methodology ....................................................................................................................... 52


4.5 Conclusion .......................................................................................................................... 62

CHAPTER 5 EDGE EFFECT .................................................................................................... 62

5.1 Introduction ........................................................................................................................ 62

viii


5.3 Methodology ....................................................................................................................... 65

5.4 Results and discussions ...................................................................................................... 66

5.5 Conclusion .......................................................................................................................... 75

CHAPTER 6 CONCLUSIONS ................................................................................................... 76

6.1 Summary ............................................................................................................................. 76

6.2 Contribution ....................................................................................................................... 77

6.3 Future research .................................................................................................................. 78

REFERENCES ............................................................................................................................. 79

CURRICULUM VITAE .............................................................................................................. 98

ix

LIST OF FIGURES

Figure 1. Karst area of China .............................................................................................. 3

Figure 2. Four-phase genetic pattern of the erosion of karst surface (after Jakucs 1977) .. 9

Figure 3. Location of the study area ................................................................................. 14

Figure 4. Distribution of fenglin/fengcong on the elevation map (Elevation unit: m) ..... 17

Figure 5. Flow chart of the methodology.......................................................................... 20

Figure 6. Example of karst landform classification .......................................................... 23

Figure 7. Verification of the classification result using Google Earth (Landscapes

surrounded by blue circles are tower karsts; landscapes surrounded by red circles are

cockpit karsts) ................................................................................................................... 24

Figure 8. Box charts of six different morphological indices (TK: tower karst, CK:

Cockpit karst) .................................................................................................................... 27

Figure 9. Location of the study area ................................................................................. 34

Figure 10. Methodology Flow chart ................................................................................. 35

Figure 11. (a) Landsat Imagery without topographic correction. (b) Landsat imagery with

topographic correction (Band combination: RGB 642) .................................................... 39

Figure 12. (a) Black shadows on the Landsat Imagery. (b) Shadows detected on the

Landsat imagery with purple highlighted (Band combination: RGB 642) ....................... 40

Figure 13. Scatterplots of NDVI and karst topography in tower karst and cockpit karst . 42

Figure 14. Windrose diagram of NDVI-Aspect on tower karst and cockpit karst ........... 42

Figure 15. The absence of vegetation at the south aspect of tower karst (Photo Image was

taken at the field survey of Guilin on Aug 2011) ............................................................. 44

Figure 16. Study area on the elevation map ...................................................................... 50

Figure 17. Annual precipitation of the study area ............................................................. 51

Figure 18. Annual temperature of the study area .............................................................. 51

Figure 19. Annual runoff of the study area ....................................................................... 51

Figure 20. Methodology of delineating of watershed ....................................................... 53

Figure 21. Hydrological chart of fengcong depression system (After Guilin karst geology,

1988) ................................................................................................................................. 54

Figure 22. Hydrological chart of fenglin plain system (After Guilin karst geology, 1988)

........................................................................................................................................... 54

Figure 23. Comparison of Annual Runoff and Annual Runoff ....................................... 55

Figure 24. Bicarbonate estimation from solute transport model....................................... 58

x

Figure 25. Soil types in the study area .............................................................................. 59

Figure 26. Distribution of pressure (h) in different depth of sample sites ........................ 60

Figure 27.Solute transport of bicarbonate ......................................................................... 61

Figure 28. Bicarbonate Concentration in the study area ................................................... 61

Figure 29.Global Moran's I for Fenglin in vertical direction ............................................ 72

Figure 30. Global Moran's I for Fenglin in horizontal direction ...................................... 73

Figure 31. Global Moran's I for Fengcong in vertical direction ....................................... 74

Figure 32. Global Moran's I for Fengcong in horizontal direction ................................... 74

xi

LIST OF TABLES

Table 1. Confusion matrix of tower and cockpit karst classification ............................... 25

Table 2. Descriptive statistics of towers and cockpits ...................................................... 26

Table 3. Digital value (DN) statistics of samples from shadow areas and non-shadow

areas .................................................................................................................................. 41

Table 4. Correlation between tower karst latitude and other variables ............................. 67

Table 5. Model summary between tower karst latitude and other variables .................... 67

Table 6. Model coefficients between tower karst latitude and other variables ................ 67

Table 7. Correlation between tower karst longitude and other variables ......................... 68

Table 8. Model summary between Tower karst longitude and other variables ................ 68

Table 9. Model coefficients between tower karst longitude and other variables.............. 69

Table 10. Correlation between cockpit karst latitude and other variables ........................ 69

Table 11. Model summary between cockpit karst latitude and other variables ................ 70

Table 12. Model coefficient between cockpit karst latitude and other variables .............. 71

Table 13. Correlation between cockpit karst longitude and other variables ..................... 71

Table 14. Model summary between cockpit karst longitude and other variables ............. 71

Table 15. Model coefficients between cockpit karst longitude and other variables ......... 72

xii

ACKNOWLEDGMENTS

I would like to express my deepest gratitude to my advisor, Professor Mick Day. The

accomplishment of the dissertation would have been impossible without his sincere help

and patient direction. Suggestions and guidance from him will become a lifetime treasure

in my future life. Special thanks also to my advisory committee members, UWM

Professors Changshan Wu, Glen Fredlund and Zengwang Xu, and Professor Daoxian

Yuan, of the Institute of Karst Geology, for their valuable suggestions and constructive

comments on my dissertation work.

I also appreciate all the people who helped me during my graduate study: Professor Mark

Schwartz, Dr. Patti Day, Professors Arthur and Margaret Palmer of the State University

of New York – Oneonta, Professor Ryan Holifield, Professor Kristin Sziarto, Dr.

Jonathan Hanes, Professor Anne Bonds, Professor Hyejin Yoon, Professor Woonsup

Choi, Dr. Alison Donnelly, the Director of UWM’s Cartography and GIS Center Donna

Genzmer, Professor Judith Kenny, and Program Associate Niko Papakis. In addition, I

also extend my thanks to former and current colleagues at UWM, Chengbin Deng, Rong

Yu, Wenliang Li, Yingbin Deng, Miao Li, Alarico Fernandes, Nick Padilla, Andrea

Kuhlman, Ashley Murray, I-Hui, Lin, Yang Song, and Wei Xu, for their support and help.

Last but not the least, my parents deserve my very special thanks for their consistent

encouragement and their support in my life.

1

CHAPTER 1 INTRODUCTION

Tower karst (fenglin) and cockpit karst (fengcong) are two fundamental representative

landscape styles in tropical karst. Previous studies have proposed both sequential and

parallel models to explain the evolution of these two karst landscapes, but neither group

of models provides a totally convincing explanation of the genesis and evolution of

existing tower and cockpit karst regions. This unclear interpretation of tower-cockpit

relationships not only obscures the understanding of the process-form dynamics in such

karst landscapes, but also confuses the definition of the two types. In addition, prior

technological limitations, as well as the fragmented nature of karst landscape, have

precluded the incorporation of much relevant geologic and other data into broad

geospatial frameworks using GIS and RS techniques, with much of the data scattered and

poorly integrated both spatially and temporally. To address these issues, this research

combines geographic, geologic and hydrologic data with the most resent GIS and remote

sensing technologies to generate integrated, novel and useful quantitative

geomorphological measures.

Specifically, there are three purposes of the research. The first is to use contemporary

GIS and RS techniques to combine hydrologic, geologic and lithologic data from

previous studies to examine the karst landscape of Guilin, China, which is arguably the

World’s best example of tower-cockpit karst. The second aim is to develop an effective

method to differentiate fengling and fengcong landforms. The third goal is to test a

hypothesis about the existence of an “edge effect” relating to fenglin and fengcong,

2

which will help to illustrate the tower-cockpit relationship, not only in the area of Guilin,

but also elsewhere.

The term “karst” is derived from the German name for the “kras” region of Slovenia,

where landscapes dominated by carbonate rock dissolution were first identified (refs).

The term has been used both to describe the processes of chemical dissolution of soluble

rocks and associated mechanical processes, such as subsidence and collapse, and to

delineate the landforms and landscapes resulting from these processes (Yuan 1998). Thus,

the term “karst” broadly includes karst processes, karst features and karst regions.

There is an apparent disparity of karst landscape distribution around the world. The

largest areas are in North America, in the Alpine fold region of Europe and in Southeast

Asia, particularly in China, Thailand and Vietnam. There are also important karst areas in

western Asia and North Africa. While karst occurs in southern Africa, South America,

Australia, New Zealand and the Pacific, it is far more extensive in the northern

hemisphere than in its southern counterpart, reflecting the worldwide distribution of

carbonate rocks (Sweeting, 1972).

Soluble carbonate rocks cover over 3.4 million square kilometers in China, which is

approximately one third of the total territory (Figure 1). Among this vast karst landscape,

the humid tropical and subtropical area of southern China possesses the most extensive

outcrop of limestones and dolostones, and exhibits the most dramatic karst landscape.

Differing from the “classical” karst areas of Europe and the more subdued karst of North

America, the karst topography of mainland China is unique from a global perspective. It

is promoted by a combination of favorable geomorphic elements: old, hard, compact

carbonate rocks, strong uplift in the Cenozoic Era, no continental ice sheet scouring

3

during the Pleistocene, and the abundance of water and heat in the East Asia Monsoon

region (Yuan, 1991). Visiting Guilin in the 1970s, Sweeting (1978) enumerated three

conditions that favored the development of karst in China: (1) large areas and great

thickness of pure limestones uninterrupted by any significant intercalation of other rocks;

(2) a warm and wet climate over a long period, uninterrupted by intense cold phases

(glaciation); (3) neo-tectonic uplift in the later phases of the Tertiary and Quaternary

periods. Configured by the Qinghai-Tibet plateau and the Qinling-Dabieshan Range,

three main karst areas can be identified on the basis of topographic and climatic setting:

the humid subtropical karst in South China, the arid and semiarid karst in North China

and the high mountain karst in Southwest China (Yuan, 1998).

Figure 1. Karst area of China

Data source: School of Environment, the University of Auckland

4

In south China, the physical strength and coherence of the carbonate rocks have a

significant influence on karst formation and landform development. In particular, at the

surface, they give rise to accentuated cockpit karst (fengcong) massifs and to spectacular

karst towers (fenglin) protruding from alluvial plains. Beneath the surface, they provide

the necessary support for large subterranean features such as underground streams and

cave chambers (Yuan, 1998). Elsewhere in the World, there is tower karst in Southeast

Asia, Central America and the Caribbean, but the towers there are generally lower and

rounded, being formed in Tertiary porous carbonate rocks (Troester, 1992).

Guilin is World-famous for its prominent tower karst (fenglin) landscape, which in terms

of tower size, diversity, concentration and extent is unequaled both in China and the

World. It is regarded as the most beautiful and iconic scenery in China (Sweeting,

1995).The karst array, especially the tower karst near the Li River, where steep cliffs

tower over the river, has created dramatic scenery that has been an inspiration to Chinese

painters for centuries. The scenery of the Guilin karst, with needle and tower-like hills of

limestone towering above the plains has greatly influenced the development of Chinese

art and made it World-famous (Swann, 1956).

There are several problems in the study of tower karst in Guilin and elsewhere in the

World. The first problem comes from the understanding of tower karst and cockpit karst.

Chinese geomorphologists distinguish two groups of tropical karst landscapes: ‘fenglin’

or ‘peak forest’ and ‘fengcong’ or ‘peak cluster’ (Zhang 1981, Yuan 1981). The former

are individual isolated residual hills rising from flood or corrosion plains and which are

related to tower karst in western terms. There is a generally accepted interchangeability

5

between the Chinese word fenglin and the western term tower karst, even though they are

defined by different parameters. Almost all true tower karst can be described as fenglin,

though the latter term does include some landscapes that contain conical hills instead of

steep sided towers (Zhu, 2005). Such interchangeability works extremely well in the case

of the residual hills, or towers, which in fenglin are simply erosional remnants in a planar

area usually dominated by allogenic fluvial drainage and an alluviated surface (Day and

Tang, 2004). On the other hand, fengcong refers to groups of residual hills sharing a

common bedrock basement and incorporating closed depressions between the clusters of

peaks – landscape to which several western terms have been applied. Waltham (2008)

insisted that the word fengcong is equivalent to the western term cone karst based on the

morphologies of the residual hills. Day and Huang (2009) proposed that cockpit karst is a

more appropriate equivalent than cone karst because of its emphasis on the active

geomorphological areas between the residual hills. Therefore, it is necessary to

distinguish fenglin and fengcong not only on the grounds of morphological differences

but also from the perspective of the terminology applied to the areas within which they

are located.

A second problem relates to the evolutionary paths of tower karst and cockpit karst, and

especially the relationship between them. This has been a long-term debate over two sets

of hypotheses and models that have proved controversial among karst geomorphologists.

Broadly, sequential models have been based upon Davis’s theory of a Geographical

Cycle (1899), and have treated tower karst and cockpit karst as different stages of

landscape evolution (Sweeting, 1958, 1990; Gerstenhauer, 1960; Song et al., 1983;

Williams, 1985; He, 1986). These models suggest that tower karst has evolved from

6

cockpit karst, and that thus the former is an “older’ stage of the latter. By contrast,

parallel models have proposed that both tower karst and cockpit karst have evolved

simultaneously without distinctive erosional stages (Verstappen, 1960, Balazs, 1968;

Yuan 1981, 1986; Williams, 1986; Yuan et al., 1990; Tan 1992; Sweeting, 1995 and Zhu

et al., 1988). Both groups of hypotheses have been mostly based on qualitative field

observation rather than detailed quantitative evidence from field measurements, and

neither of them appear to fully explain the origin and development of the tower karst in

the Guilin area. Perhaps surprisingly, the fengcong karst in Guilin actually occupies a

larger distribution area than does the fenglin karst (Zhu, 2005). In part, this has led to the

suggestion that there is an “edge effect” (Day and Huang, 2009), with fenglin developing

around the periphery of fengcong, and this suggestion offers a fresh perspective from

which to investigate whether fenglin has evolved from fengcong or whether they have

parallel evolutionary paths.

The third problem relates to spatial scales, technological changes and data integration.

Previous studies of cockpit and tower karst topography have focused on individual and

assembled morphology and classification, and have used these, largely qualitative

descriptions of small-scale areas as the basis for landscape evolution interpretation. In the

context of Guilin, Sweeting (1978) noted that the sequential model, which dominated the

thought of Chinese geographers, led them to divide the landscape of Guilin into three

different stages: fengcong, fenglin and dufeng (isolated peaks), without any quantitative

basis. In addition, due to technological limitations and the fragmental nature of the karst

landscape, geomorphologists did not adequately combine material (geological) and

7

geomorphological process data into a broad geospatial frame, with previous studies being

disparate in focus and scope.

With the rapid recent development of RS and GIS technologies, it is important to apply

such technologies to determine how they can advance geomorphological research on

tower and cockpit karst. Detailed GIS datasets and increasing spectral resolution of

remotely sensed imageries should provide a deeper insight into, for example, identifying

individual tower and cockpit features in the Guilin karst and non-karst landscapes, and

incorporating such spatial data with recent hydrological evaluations. This will allow for

testing of such hypotheses as that concerning the “edge effect”.

CHAPTER 2 DIFFERENTIATING TOWER KARST (FENGLIN) AND COCKPIT

KARST (FENGCONG) USING DEM CONTOUR, SLOPE AND CENTROID

2.1 Introduction

Tower karst (fenglin) consists of isolated limestone towers/hills, often with vertical flanks

rising from alluvial plains (Zeng 1982; Day and Tang 2004). Cockpit karst (fengcong),

on the other hand, involves similar dimension enclosed depressions surrounded by

overlapping hills and ridges. (Day 2004a; Day and Chenoweth 2004; Yuan 1984; Zhu et

al. 2013). These two landforms are the two most spectacular and diagnostic landscape

styles in tropical karst environments, where high temperatures and abundant precipitation

provide a favorable environment for rapid and prolonged corrosion. Typical examples

can also be found in Southeast Asia, Central America and the Caribbean, although the

8

towers there are lower and rounded, being formed in weaker, porous Tertiary carbonate

rocks (Troester 1992).

Studying the relationships between fenglin and fengcong holds the key to understanding

tropical karst evolution (Sweeting 1972; Smart et al. 1986; Zhu 1988; Yuan 1991 and

2004; Ford and Williams 2007; Waltham 2008). Jakucs (1977) proposed a four-phase

genetic model (Figure 2) to describe the erosion of a karst surface: (1) soil and regolith

are removed and accumulate in karst depressions, (2) tectonic uplift occurs as the karst

plain becomes corroded with cockpit karst, (3) as the uplift ceases or decelerates,

floodplains widen and isolate cockpit karst, (4) when residual blocks widen at the level of

water table, they are eventually isolated as towers or tower groups. Yuan (1985) reviewed

explanations proposed for the evolution of tower karst, among which are two extreme

possibilities: (1) tower karst evolves directly and independently of any previous

morphology, given favorable lithological, relief and climatic circumstances, (2) the

geometry of tower karst depends explicitly on the topographic characteristics inherited

directly from a previous phase, such as cockpit karst relief, but becomes increasingly

independent of this inheritance as erosion proceeds.

9

Figure 2. Four-phase genetic pattern of the erosion of karst surface (after Jakucs 1977)

10

There has been much debate as to whether the two styles follow sequential or

simultaneous development paths (Williams 1987; Zhu 1988; Yuan 2004; Waltham 2008),

although the consensus now appears to be that the two forms can and do develop

contemporaneously, with fenglin, which in many locations occurs on the periphery of

fengcong, involving greater fluvial influence (McDonald 2002; Tang and Day 2000; Day

2002; Day and Huang 2009).

The distinction between the two styles is not clear-cut, however, and their morphological

similarities and spatial coexistence make it difficult in some locations to differentiate

between them, particularly where local conditions result in intermediate forms that fit

neatly into neither category. For example, some residual hills or towers in fenglin share

common bases, and flat-floored depressions may isolate individual fengcong hills.

Terminology apart, both tower and cockpit karst contain assemblages of residual hills of

variable morphology, some of which are connected by ridges, others of which are

isolated by equally variable depressions (Day 1978, 1981 and 2004a).

Initial approaches to differentiating between cockpit karst and tower karst focused on

the morphology and spatial arrangement of the residual hills, which are a common

denominator of both styles. Balazs (1973) assigned the hills to four “type area” categories

on the basis of their 𝑅𝑑/ℎor diameter/height ratio: Yangshuo (D/H <1.5), Organos (D/H

1.5-3.0), Sewu (D/H 3-8) and Tual (D/H >8), suggesting that the relatively high and

steep hills were more typical of fenglin, although they might be in a minority even here, a

suggestion confirmed in a tower karst area of Puerto Rico, where the Yangshuo type

accounted for only 2% of the hills measured (Day 1978). While residual hill

11

morphologies may provide some insights into the development of tower and cockpit karst,

clearly they provide no distinction between the two styles, except possibly when

combined with the spatial arrangement of the hills and the intervening depressions, for

example to derive a generic three-fold classification of tropical karst landscape styles:

Type I, in which enclosed depressions dominate and are interspersed with subdued hills;

type II, in which enclosed depressions and residual hills attain approximately equal

prominence; and type III, in which isolated residual hills dominate intervening near-

planar surfaces (Day 1978 and 1981). Following this approach, fengcong/cockpit karst

belongs to type II, whereas tower karst belongs to type III.

Another approach attempted to differentiate tower and cockpit karst on the basis of hill

summit elevations (Gellert 1962), but statistical analysis of over 600 summits in the

Guilin area demonstrated that there was no distinct correlation between summit

elevations and styles (Zhu et al. 1980). Furthermore, it is clear that the terms tower karst

and cockpit karst actually refer to spectra of landscapes, with there being, for example,

four major types of tower karst: (1) residual hills protruding from a planed carbonate

surface veneered by alluvium; (2) residual hills emerging from carbonate inliers in a

planned surface cut mainly across non-carbonate rocks; (3) carbonate hills protruding

through an aggraded surface of clastic sediments that buries the underlying karst

topography; (4) isolated carbonate towers rising from steeply sloping pedestal bases of

various lithologies (Ford and Williams, 2007).

Although residual hills (and ridges) are integral to both cockpit and tower karst

landscapes, the dynamic foci of geomorphic processes are the areas between them:

sinkholes of variable morphology, erosional valleys, flat-floored depressions or alluviated

12

plains. Their role is critical in karst style differentiation, and ultimately it is the

relationships between the residuals and the intervening depressions that distinguish

between cockpit karst and tower karst. Broad morphometric analysis of polygonal

tropical karst has been successful in assessing overall patterns of symmetry and otherwise

(Williams, 1972, Day, 2004b), and using depth/diameter and width/length ratios of

adjacent closed depressions and residual hills revealed nuances within type II cockpit

karst landscapes (Day 1978, 1982), so similar approaches have promise in fengcong-

fenglin differentiation. Fourier series analysis of Jamaican karst showed that terrain

wavelengths could be used to uncover patterns of organization, including geologic

influence, and distinguish between doline (type I) and cockpit (type II) landscapes (Brook

and Hanson, 1991). Broadly, surface roughness may serve as a useful discriminator (Day,

1979, 1981; Day and Chenoweth, 2013a and 2013b), and differences in other variables,

such as cave density, may also help in tower-cockpit differentiation (Zhu, 1982; Zhu et al.

1988), but these are not considered further here.

Despite the fact that utilizing DEMs and remote sensing to analyze landforms is well

established in other branches of geomorphology (Bolch et al. 2005; Bubenzer and Bolten

2008), the application of DEMs in tropical karst geomorphology has remained

challenging and immature due to unsolved issues. One issue is that the rugged

topography blocks direct solar radiation, so satellite images are embedded with deep

shadows, which cause image inconsistency because covered areas display reduced

reflectance value compared to non-shadow areas with similar cover characteristics (Giles,

2001). Remote sensing software, such as ArcGIS and ERDAS IMAGINE tends to

classify the shadow areas as a different category even if they have cover characteristics

13

similar to un-shadowed areas, such that the results of preliminary image classification do

not represent the actual geometry of tower and cockpit karst. These issues have been

addressed by topographic correction (Zhang et al. 2011) and spectral mixture analysis

(Yue et al. 2010), which have achieved mitigation yet were time consuming and labor

intensive in terms of cost and effect.

Another issue of geomorphologic mapping is that karst features with similar

geomorphologic characteristics but different geologic histories may easily be

misclassified (Ho 2011; Mylroie and Mylroie 2009; Purkis et al. 2010). Karst landforms

can cause more confusion than those in other geological settings, and so image

classification of surface karst landforms via remote sensing software should ideally be

combined with field survey data (Day, 2004b).

Measurement of tower and cockpit karst is crucial to understand their morphological

differences and to distinguish their evolutionary paths on a geochronological scale, and

such measurement ultimately requires the use of RS and GIS (Day, 2004b). Digital

Elevation Models (DEM) are critical in this context, playing a “…fundamental role in

modulating earth surface and atmospheric processes” (Hutchinson and Gallant,2000, p29).

Remotely-sensed data, Digital Elevation Models and GIS techniques have been used to

delineate karst features in the Cockpit Country of Jamaica (Chenoweth and Day 2001;

Lyew-Ayee 2004; Lyew-Ayee et al. 2007; Fleurant et al. 2007, 2008 ) but these

techniques have yet to be employed in discriminating between cockpit and tower karst.

The aim of this chapter is to classify tower karst and cockpit karst with greater accuracy

and less confusion than hitherto, and this effort was attempted developing a novel method

to categorize the two landscapes by utilizing the contours, slope and centroid derived

14

from the ASTER GLOBAL DEM (Digital Elevation Model). Classification accuracy was

assessed through the verification of the corresponding region of interest exported on high

resolution satellite imagery of Google Earth as the reference. Morphological indices were

used to compare and contrast geomorphic variations using Object Based Image Analysis

(OBIA).

2.2 Study area and data

The study area is located near Guilin, in the Guangxi Autonomous Region of China

(Figure 3), which is renowned for its spectacular tower and cockpit karst landscape, and

where the development of the two landscape styles is promoted by a unique combination

of climatic, hydrological and geological conditions (Yuan, 1991, 2004; Zhu 1988).

Figure 3. Location of the study area

15

According to the Guilin Karst and Geological Structure report (1988), the geological

framework of Guilin involves distinct basement and cover structures. The basement is

characterized by NE-extending synclinoria composed of lower Paleozoic rocks, while the

overlying cover structure is made up of Devonian and Carboniferous rocks, with

Indosinian N-S-extending arcuate structures as the main framework, upon which are

superimposed early Yanshanian NNE-trending Neocathaysian structures and late

Yanshanian NW-trending structures. The protoliths of these tectonites are all carbonate

rocks. Owing to the activity of karst hydrology in the later stages, the cementing material

is often replaced by calcareous and argillaceous substances, forming

metasomatictectonites.

Under the influence of monsoons from the Indian and Western Pacific Oceans, the study

area has pronounced dry and wet seasons, and 80 to 90% of total annual precipitation is

received from May to October (Zhao, 1986). The climate in Guilin is a subtropical

monsoon humid type (Liu 1991), with an annual average precipitation of 1873.6 mm and

annual average temperature of 18.8℃ (Yuan, 1992).

The major drainage system is that of the Li River, with both significant allogenic

surface drainage and autogenic ground water contributions. The approximate catchment

area upstream of Yangshuo is 5520 km2 (Ru et al. 1988), with the estimated annual

average recharge from the adjacent non-carbonate area being 41.9 x 108 m3/year, and the

average discharge flowing out from the basin being 67.8 x 108 m3 yr-1, giving an average

regional recharge of 25.9 x 108 m3 yr-1 (Huang et al. 1988).

Spatially extensive, thick, pure and vertically uninterrupted Paleozoic age carbonate

rocks provide the material base for the development of tower and cockpit karst in the

16

Guilin area (Zhu et.al. 1988), within a basin surrounded by distinct non-karst uplands:

Yuechenglin to the north, Haiyangshan to the east, and Jiaqiaoling to the southwest.

Within the karst landscape (Figure 4), tower karst accounts for 47% of the area and

cockpit karst 53% (Zhu, 1982).

17

Figure 4. Distribution of fenglin/fengcong on the elevation map (Elevation unit: m)

18

2.3 Methodology

2.3.1 Main procedures

For geomorphological mapping, the ASTER Global Digital Elevation Model (GDEM)

with 30 m spatial resolution was acquired from the website

http://gdem.ersdac.jspacesystems.or.jp/. Related GIS data sets were acquired from the

National Institute of Karst Geology in Guilin. All images and GIS data were projected

with zone 49N and WGS 1984 datum.

Before implementing the model to classify tower karst and cockpit karst, several

variables, including contour, centroid and slope were selected. Contour line delineates the

actual hill profile at a certain altitude and can be used to identify the two landforms.

Centroid reveals the number and position of the geometric center within the contour

polygons, and therefore can be used to differentiate the tower and cockpit features. Slope

represents the steepness of the hill and can eliminate unmatched hills with an appropriate

threshold applied.

The ASTER GDEM data was first subset to the sampling area, and contour lines were

extracted from the DEM at 10, 20, 30, 40 and 50 m intervals. Slope data was derived

from the DEM and areas of >30 degrees were extracted because these are typical of tower

and cockpit karst (Zhu, 1982). In order to use the contour lines which best represent the

actual landforms, those with the smallest linear distance to the outer rim of the 30 degrees

slope were selected. The closed contour lines were then converted to polygons, and the

centroids of the polygons were then extracted. In differentiating tower and cockpit karst,

19

polygons of the former contain only one centroid, whereas cockpit polygons contain

multiple centroids. After initial landform classification, the classification shape results

were exported as KML files, and these files were referenced to Google Earth to assess

how well the results match with the landforms on the high resolution imageries. Detailed

procedures are displayed in Figure 5.

20

Figure 5. Flow chart of the methodology

Contour line

extracting

Conversion of closed

contour lines to polygons

Morphological

measurement

Exported as KML

files

Accuracy

assessment

Areal subsetting of

ASTER DEM

>1

=1

Centroid extracting

Classified as tower

karst

Classified as cockpit

karst Centroid

counting

Slope calculation and

reclassification

Selection of contour lines

based on matching to the

outer rim of 30 degree

slopes

Classification

results

21

2.3.2 Morphological measurement

After accuracy assessment, the classification results of shape files containing the

original karst polygons were first converted to GeoTIFF format and then introduced into

Trimble’s eCognition Developer 64 8.7 software package to conduct object-based image

analysis (OBIA). Image segmentation was performed by adopting 25 for the scale level

and 0.5 for the color level. After segmentation, the new polygons generated by the

software were checked and any sub-divided polygons sharing borders were manually

merged with each other in order to maintain their shape identical to the original polygons.

Area, length, length/width ratio, main direction, roundness and shape index were selected

from the export result setting, then the new polygons containing the morphological

attributes were exported.

In eCognition 8.7 software, both the asymmetry and length/width ratio feature describe

the relative length of an image object compared to a regular polygon by approximating an

ellipse. Since the asymmetry was less accurate than the length/width ratio (Trimble 2011),

the ratio index was used for calculation. The main direction (the angle between the longer

axis of polygons and North) was added to the polygon attributes to measure the direction

of polygon orientation. In addition, the shape index (the ratio of the perimeter of the

actual landform to the perimeter of a circle with the identical area) was also calculated to

check the smoothness of the generated polygons.

22

2.4 Results and Discussion

2.4.1 Accuracy assessment

The accuracy assessment was based on the classification results, with Figure 6 showing

an example from the sampling area after classification. One hundred polygons were

selected randomly for tower karst and cockpit karst respectively and exported to Google

Earth for validation (Figure 7).

23

Figure 6. Example of karst landform classification

24

Figure 7. Verification of the classification result using Google Earth (Landscapes

surrounded by blue circles are tower karst; landscapes surrounded by red circles are

cockpit karst)

A confusion matrix (Kohavi and Provost, 1998) was used to report the classification

accuracy (Table 1). The matrix contained actual results from Google earth and classified

results from the proposed model. The column of tower and cockpit represent the number

of correct and incorrect results in the corresponding category in the actual class, while the

row of tower and cockpit represent the instances in the classified class. The producer’s

accuracy is calculated from the proportion of correct results in a specific column for the

actual class. For example, the producer’s accuracy for tower in the matrix is 74 out of

(74+14), reaching 84.09%. Likewise, the user’s accuracy is calculated from the

proportion of correct results in a specific row for the classified class. As for the overall

accuracy, it is calculated from the proportion of correctly classified results for both tower

25

and cockpit in the total of 200 sample points. The result indicated that, the accuracy

measurement for both tower karst and cockpit karst attained better than 70% success in

both the user’s and producer’s accuracy categories. The user’s accuracy for cockpit was

86 %, which was higher that the tower. The relatively lower accuracy for tower is due to

the geographical difference between tower karst and cockpit karst. Most cockpit karst in

Guilin is located in rugged mountain areas with higher elevation and intervening

depressions. By contrast, most tower karst is concentrated in the flat plain with lower

elevation and relief with some sporadic distribution in the rugged mountain areas.

Therefore, tower karst on the lower flat plain can be correctly classified by the contour

lines with the lowest interval of 10 m, while those located in rugged mountain area are

captured by the contour lines with interval larger 10 m and thus misclassified as cockpit

karst. Another reason that may contribute to the misclassification is the mixture of tower

and cockpit karst within the same area. Nevertheless, the overall accuracy in the

sampling area reached 80%.

Table 1. Confusion matrix of tower and cockpit karst classification

Classified

Actual

Tower Cockpit User’s accuracy (%)

Tower karst 74 26 74

Cockpit karst 14 86 86

Producer’s accuracy (%) 84.09 76.79 80

2.4.2 Morphological measurements

The results of the morphological measurements are shown in Table 2 and Figure 8,

which indicate several aspects of geomorphic variation between tower karst and cockpit

karst. First, the overall areas and the total perimeters of cockpit karst exceed those of

tower karst. Second, the tower karst displays greater circularity than the cockpit karst, as

26

reflected in Figure 8, where roundness values in the tower karst are predominantly in the

range between 0 and 0.5. Additionally, the longer axes of most tower features are

oriented NE-SW to E-W, whereas the cockpits are mostly oriented in a broader arc NE-

SW to SE-NW. Third, based on the comparison of shape indices, the towers are smoother

shaped than the cockpits, with the cockpits having greater shape irregularity. Finally, the

length/width ratios of both towers and cockpits is remarkably similar.

Table 2. Descriptive statistics of towers and cockpits

Tower karst

Area (sq.m) Perimeter

(m)

Length/Width Main

Direction

(degree)

Roundness Shape

index

Mean 33614.779 887.390 1.630 87.665 0.455 1.315

Standard

Deviation

32216.878 483.897 0.468 46.849 0.274 0.142

Minimum 162 54 1 0.430 0 1.061

Maximum 161919 2538 3.251 176.884 1.807 1.7321

Cockpit karst

Mean 272949.674 4349.558 1.685 96.348 0.838 1.613

Standard

Deviation

240668.444 4176.807 0.495 48.267 0.433 0.381

Minimum 196 56 1 0.730 0 1

Maximum 1006656 28336 3.367 177.284 1.731 2.933

27

Figure 8. Box charts of six different morphological indices (TK: tower karst, CK:

Cockpit karst)

2.4.3 Implication for geomorphologic classification and morphometric measurement

Liang and Xu (2013) utilized DEM to extract several variables such as area ratio,

elevation, slope, and curvature to differentiate tower karst and cockpit karst in Guilin.

They used a previous survey map as the reference to validate their classified results. The

variables they selected are predominantly the most basic morphometric indices that can

be easily extracted from DEM. Although these indices can differentiate tower and cockpit

karst to some extent, they failed to provide detailed information about the spatial

28

distribution and dimension of the two landforms. On the other hand, their method of

sampling and validation, although it initially sounds clear and correct, is problematic for

several reasons. First, the samples collected from different tiles, no matter what scale they

represent, are actually mixtures of tower and cockpit to some extents. Second, the

reference map only provides general location but not accurate position of tower and

cockpit karst. Third, the landforms have gone through a considerable change given the

fast development of Guilin in the past decade, while the survey map was based on a

survey conducted many years ago. Therefore, samples for morphometric analysis are

not pure tower or cockpit type, and the survey map cannot provide reliable and accurate

information for validation.

In contrast, the approach here not only explicitly classifies the tower and cockpit karst,

but also delineates their actual outlines. As for the validation method, the most updated

satellite imagery on Google Earth was used as the reference to validate the classified

results. In addition, the object-based image analysis also produced accurate

measurements of the two landforms that can reflect their morphometric and spatial

properties.

2.5 Conclusion

Geomorphic classification of tower karst (fenglin) and cockpit karst (fengcong) using

GIS and RS has been problematic because of shadows on the imagery. Topographic

correction and spectral mixture analysis have provided some mitigation but they are time

consuming, labor intensive, and incomplete. Here a combination of the contour lines,

slopes, and centroids derived from ASTER DEM has been used to differentiate the karst

landforms. The results suggest that using the DEM rather than optical sensor imagery in

29

karst surface extraction is a feasible way to avoid shadow problems and misclassification.

This approach is also cost effective compared to topographic correction and spectral

mixture analysis.

The extraction of morphological parameters of towers and cockpits was based on object-

based image analysis, which provided a rapid and semi-automatic way to calculate

geomorphic attributes of the two landform classes. Morphological analysis indicates that

the towers and cockpits have significant morphometric variations in terms of area,

perimeter, roundness, and shape. Main directions and length/width ratios are similar,

perhaps warranting further study from a hydrogeologic perspective.

CHAPTER 3 INVESTIGATION OF VEGETATION ON SURFACE KARST

USING TOPOGRAPHIC CORRECTION, SHADOW RETRIEVAL AND

VEGETATION INDEX

3.1 Introduction

Karst landscape represents a fragile and heterogeneous environment constrained by local

geology (Yuan and Zhang 2008; Parise, Wales, and Gutierrez 2009). Through the rapid

dissolution of soluble rocks such as limestone and dolostone, topical karst develops

highly diversified surface landforms and subsurface caves. Among the different tropical

karst geomorphological types, tower karst (fenglin) (Day and Tang, 2004) and cockpit

karst (fengcong) (Day, 2004; Day and Chenoweth, 2013) are the two most spectacular

and their interrelationships are critical to the understanding of tropical karst evolution as

well as other surficial processes (Sweeting, 1972; Smart et al., 1986; Zhu, 1988; Yuan,

1991, 2004; Ford and Williams, 2007; Waltham, 2008).

30

Tropical karst landscapes, including those in tower karst and cockpit karst, are composed

of a variety of land covers, including soil, bedrock, water, and vegetation. Extensive

exposures of bedrock, however, are not a major land cover type in tropical karst area

unless karst rocky desertification occurs in the appearance of a stone-desert-like

landscape (Wang et al., 2004; Want and Li, 2007). By contrast, vegetation can be

classified as photosynthetic vegetation (PV) and non-photosynthetic vegetation (NPV).

The former category includes plants that have green leaves, and the latter is referred as

the plants that lack significant amount of chlorophyll (aboveground dead biomass, litter

and wood) (Guerschman et al., 2009; Roberts et al., 1993).

Research into the linkages between vegetation and geomorphology, especially on

hillslopes has developed significantly since the 1950s (Marston, 2010), but the focus has

been unidirectional until recently (Viles, 1988). On the one hand, botanists and landscape

ecologists tended to view vegetation as responsive to geographic processes and focused

on the effect of topography on vegetation (Reinhardt et al., 2010). On the other hand,

geomorphologists tended to treat vegetation as an independent variable that affects

landforms and sediment at limited spatial–temporal scales (Renschler et al., 2007). Others

(Harden, 2002; Marston et al., 2003; Keesstra et al., 2009), however, began to view the

relationships between vegetation, geomorphology and landforms as dynamic, although

the responses of vegetation to geomorphic processes need greater attention (Marston,

2010).

Traditional methods rely on field surveys of tropical karst vegetation type, structure, and

on geomorphic measurement of hillslopes to assess association between vegetation and

landform morphology. These methods are labor intensive and time consuming, making it

31

difficult to apply them at large geospatial scales (Wang and Li 2007). By contrast,

remotely sensed data and related analysis techniques provide excellent data source at

broad geographic scale and represent a cost effective way to extract vegetation cover

information (Asner and Heidebrecht, 2003; Asner et al., 2005). Such approaches have

been widely used in recent vegetation studies (Kim and Daigle 2011; Yue et al, 2013).

Vegetation indices have their own strengths and drawbacks in retrieving vegetation

information. Indices such as the Normalized Difference Vegetation Index (NDVI) utilize

spectral contrasts between chlorophyll absorption in the visible-red wavelength and

cellulose scattering in the near-infrared wavelength (Tucker, 1979; Ustin et al., 2004) to

identify green vegetation. Although the indices are not intrinsic bio-physical quantities of

vegetation, they are widely used to assess vegetation vigor. However, their performance

and suitability are place-dependent and determined by the sensitivity of the index to the

characteristics of interest (Haboudane, Miller, and Pattey 2004). Nonetheless, NDVI is

one of the most widely used methods of extracting vegetation information in most cases.

One significant issue with remote sensing of vegetation in tropical karst areas,

particularly rugged ones, is the existence of shadows, which are cast on the ground,

particularly on the sides of the hills due to the differentiation of direct illumination

between sunny and shady slopes (Giles, 2001; Salvador et al.,2001; Yao and Zhang,

2006). Shadows not only cause reduction of the spectra of the shaded objects, but they

also may lead to underestimation or misclassification of land cover classes (Dare, 2005;

Hodgson et al. 2003; Weng, 2012).

There are several approaches to alleviate the influence of shadows on satellite imagery.

Shadow detection and shadow restoration are considered in shadow correction algorithms.

32

For shadow detection, this process relies on the calculation of solar elevation, solar zenith

and use of the Digital Elevation Model (DEM). In general, such algorithms include two

basic types: thresholding and modelling (Liu and Yamazaki, 2012). To date, most shadow

recognition techniques are based on setting a threshold value of the digital number such

as a histogram to differentiate shadow regions from non-shadow regions. Pixels with

digital number values smaller than the specific threshold are classified as shadow areas,

while those higher than the threshold are classified as non-shadow areas. Modelling

techniques, on the other hand, rely on specific mathematic concepts with related

topographic information to simulate shadow regions (Shahtahmassebi et al. 2013). As an

alternative approach, the Continuum Removal Technique was also applied in several

studies (Zhou et al. 2014, Huang et al. 2004). This can normalize the spectra and allow a

comparison of individual absorption from a common baseline (Kokaly, 2001). Several

techniques have been proposed for removing shadows from the satellite imagery (Yang et

al., 2007; Gao and Zhang, 2009). The most common approaches employ band ratio and

vegetation indices (Riaño et al., 2003; Mather, 2004; Yesilnacar and Suzen, 2006; Jensen,

2007). Unfortunately, due to the loss of spectral resolution (Riaño et al., 2003), band

ratios are nonlinear and subject to additive noise effects (Mather, 2004; Jensen, 2007).

Alternative efforts in shadow restoration have been made using radiometric enhancement

such as gamma correction (Nakajima et al., 2002), object-based approach (Zhan et al.,

2005), linear-correlation (Sarabandi et al., 2004; Chen et al., 2007) and histogram

matching (Sarabandi et al., 2004; Dare, 2005; Tsai, 2006). These methods all have their

strengths and drawbacks as well. For instance, gamma correction uses a single gamma

parameter for all pixels and hence ignores the existence of different backgrounds of

33

shadow areas (Zhan et al., 2005). Histogram matching can recover the digital values of a

shadow region by matching its histogram to that in non-shadow region, but this approach

is sensitive to the window size of the matched histogram (Shahtahmassebi et al 2013).

Few studies have adopted a comprehensive approach to shadow detection, vegetation

restoration and vegetation index in rugged tropical karst terrains, so they have been

unable to evaluate how vegetation correlates with hillslope topography. In many studies,

shaded areas are left unclassified or simply classified as shadows (Shackelford and Davis,

2003), leading to misclassification of land cover information. Therefore, it could be

advantageous to combine the aforementioned techniques together to investigate the topic.

The aim of this chapter is to 1) Explore the potential of combing topographic correction

and shadow restoration in the tower karst (fenglin) and cockpit karst (fengcong) of Guilin,

China; 2) Calculate vegetation indices of tower/cockpit karst using NDVI and 3) Analyze

the correlation between vegetation and topographic properties of tower karst and cockpit

karst.





of climatic, hydrological and geological conditions (Yuan, 1991, 2004; Zhu 1988). A

scene of cloud free Landsat ETM+ imagery on Oct 30, 2000 was acquired from the

website of the Global Land Cover Faculty, University of Maryland (www. glcf.umd.edu).

An ASTER Global Digital Elevation Model with 30 meters resolution was acquired from

34

Earth Center (http://gdem.ersdac.jspacesystems.or.jp/). All data were projected using

zone 49N and WGS 1984 datum.

Figure 9. Location of the study area

3.3 Methodology

3.3.1 Main procedure

The ASTER GDEM and the LANDSAT data were first subset to the study area, and the

C-correction method (Teillet et al.1982) was applied for topographic correction. Shadow

detection was achieved by using continuum removal, then shadow restoration was

completed by collecting sample points from non-shadow aspects and using them to

restore those samples on the shadow aspects. Classifying methods and results are based

on previous study by Huang et al. (2014). The NDVI was then calculated based on the

data after topographic correction and shadow restoration. NDVI and derived DEM

products of elevation, aspect and slope were combined to analyze the correlation among

them. The changes of Digital Number (DN) value in shadow areas before and after

35

correction were also compared to verify the change of NDVI value. Detailed procedures

are displayed in Figure 10.

Figure 10. Methodology Flow chart

Calculation of NDVI

NDVI

Topographic correction

(C-correction)

Shadow restoration of the

north aspect

ASTER DEM

LANDSAT

Correlation analysis

between vegetation

and karst topography

Shadow detection

(Continuum Removal)

Sampling of the north

(shadow) and south

(non-shadow) aspect

Topographic analysis

of DEM

Classification

results of

tower karst

and cockpit

karst

Elevation,

Aspect, Slope

Comparison with

field observation

36

3.3.2 Topographic correction

In order to mitigate the influence of rugged topography, topographic correction was

implemented by using the C-correction method (Teillet et al.1982), which is based on the

most widely used cosine correction (Teillet et al.1982) with the assumption that the

surface is Lambertian characteristic and capable of being a perfect diffuse reflector. For

this reason, cosine correction can only correct illumination differences caused by

orientation of the surface (Jones et al., 1988).

𝐿𝑛 = 𝐿cos 𝜃

cos 𝑖

cos 𝑖 = cos 𝜎 ∗ cos 𝜃 + sin 𝜎 ∗ sin 𝜃 ∗ cos(𝛽 − 𝜔) (3.2)

Where 𝐿𝑛 is the normalized reflectance, 𝐿 is the uncorrected reflectance, 𝜃 is the solar

zenith angle, 𝑖 is the solar incidence angle, 𝜎 is the slope angle, 𝛽 is the aspect angle, and

𝜔 is the solar azimuth angle.

The C-correction method is an extension of cosine correction by introducing the constant

of C. There is a linear relationship between 𝐿 (uncorrected reflectance) and cos 𝑖 (cosine

of the solar incidence angle).

𝐿 = 𝑎 + 𝑏 cos 𝑖 (3.3)

𝐶 =𝑎

𝑏

Where 𝑎 is the intercept and 𝑏 is the regression slope.

The C-correction can be expressed as the following equation:

𝐿𝑛 = 𝐿cos 𝜃 + 𝐶

cos 𝑖 + 𝐶

(3.1)

(3.4)

37

3.3.3 Shadow detection

Shadow detection was achieved by applying a continuum removal method (Huang et al.

2004) to identify shadows that cannot be removed from previous topographic correction.

The continuum removal reflectance (𝑅𝑐𝑟) can be calculated by dividing original actual

reflectance value at a specific wavelength (𝑅) for each band (𝜆𝑖) by the reflectance value

(𝑅𝑐) of the continuum line (convex hull) at the corresponding wavelength.

𝑅𝑐𝑟(𝜆𝑖) =𝑅𝜆𝑖

𝑅𝑐(𝜆𝑖)

The image was first processed with continuum removal, then the shadows were extracted

and exported for the next procedure.

3.3.4 Shadow restoration

In order to extract vegetation signature from karst hills, it is necessary to consider the

topographic effect, specifically the shadows cast on the north sides where direct sunlight

was blocked by the sunlit side of the hill. Each karst hill in the study area can be divided

into shaded areas and non-shaded areas (sunlit slope). Using results from a previous study

(Huang et al. 2014), samples were first collected from both sun-facing aspects and sun-

shaded aspects. Then, mean and standard deviation of the pixels of each sample points

were calculated. Digital values of sun-shaded slopes in the study area were corrected with

the linear-correlation method (Nakajima et al., 2002; Sarabandi et al. 2004; Zhan et al.,

2005; Chen et al., 2007).

𝐷𝑁𝑟𝑒𝑐𝑜𝑣𝑒𝑟𝑒𝑑 =𝜎𝑛𝑜𝑛−𝑠ℎ𝑎𝑑𝑜𝑤

𝜎𝑠ℎ𝑎𝑑𝑜𝑤

(𝐷𝑁𝑠ℎ𝑎𝑑𝑜𝑤 − 𝜇𝑠ℎ𝑎𝑑𝑜𝑤) + 𝜇𝑛𝑜𝑛−𝑠ℎ𝑎𝑑𝑜𝑤

(3.6)

(3.5)

38

Where 𝜇 is the mean value and 𝜎 is the standard deviation.

3.3.5 Deriving the topographic indices from DEM

Topographic indices such as elevation, aspect and slope were derived from the Digital

Elevation Model (DEM) respectively by using surface tools within the spatial analyst

tools of ArcToolbox in ArcGIS. The aspect is an indicator of slope direction that

represents the maximal change of rate downslope direction from each cell to its neighbors.

The value of aspect is measured from North (0 degrees) to a clockwise round of due

North (360 degrees).

3.3.6 Correlation analysis between corrected NDVI and karst topography

NDVI was calculated before and after image correction. Then, correlation analysis was

conducted in order to investigate the relationship between corrected NDVI and

topographic indices (elevation, aspect and slope) of tower karst and cockpit karst.


3.4.1 Result of topographic correction

Topographic correction using the C-correction method showed that shadows can be

alleviated or even eliminated in lower relief regions, but in regions with greater local

relief, shadows are still pronounced and need to be corrected (Figure 11).

(3.7)

39

Figure 11. (a) Landsat-7 Imagery without topographic correction. (b) Landsat imagery

with topographic correction (Band combination: RGB 642)

3.4.2 Result of Shadow Detection

The image was first processed with continuum removal, then the shadows were extracted

and exported for the next procedure. The continuum removal technique detected most

shadows and their shapes and positions match well with the actual presence on the image

(Figure 12).

40

Figure 12. (a) Black shadows on the LANDSAT-7 Imagery. (b) Shadows detected on the

LANDSAT-7 imagery with purple highlighted (Band combination: RGB 642)

3.4.3 Results of Shadow Restoration

Exactly 74,707 samples were collected from the shadow region on the north aspects and

the same number of samples were collected from a 100-meter buffer of the north aspect

within the south aspects. Mean and standard deviation of Band 4, Band 3 and NDVI

before and after linear correction are calculated, respectively.

41

Table 3. Digital value (DN) statistics of samples from shadow areas and non-shadow

areas

Mean Standard

Deviation

Shadow areas

Band 4 before

correction

27.049 12.572

Band 4 after

correction 57.306 18.298

Band 3 before

correction

28.264 5.424

Band 3 after

correction 38.054 8.578

NDVI before

correction

-0.056 0.141

NDVI after

correction

0.196 0.138

Non-shadow

areas

Band 4 57.306 18.298

Band 3 38.054 8.578

NDVI 0.178 0.142

The advantage of using linear correction of DN values in shadow areas is that the

corrected values achieved the same mean and standard deviation as those values from

non-shadow areas (Table 3). In addition, Table 1 also showed that mean value of NDVI

in shadow areas have been improved from -0.056 to 0.196, a level close to that in non-

shadow areas (0.178). This will place the appropriate values for those dark pixels on the

original NDVI image and provide better data for further analysis.

3.4.5 Comparison of the NDVI with topography between tower karst and cockpit

karst

42

Figure 13. Scatterplots of NDVI and karst topography in tower karst and cockpit karst

Figure 14. Windrose diagram of NDVI-Aspect on tower karst and cockpit karst

43

The correlations between vegetation coverage and karst topography were conducted

based on the 128 sample points from both tower karst and cockpit karst (Figure 13). The

low value of R square in all scatterplots suggests insignificant correlation between

vegetation and variable topographic parameters. However, there is a clear distinction

between the vegetation on tower karst and those in cockpit karst. For elevation, the

distribution of vegetation for tower karst ranges from the 200m to 450m level (absolute

elevation level) and such distributions are dispersed. By contrast, the distributions of

vegetation for cockpit karst are relatively concentrated at the 400m-420m level with a

slightly positive correlation. For aspects (Figure 14), vegetation on tower karst is more

predominantly distributed in the north-east orientation (0-100 degree) and sporadically

dispersed around 150 degrees. In particular, there is a pronounced absence at the south-

south west direction (150-200 degrees), which is remarkable compared the tower karst in

the study area (Figure 15). In the case of cockpit karst, the absence of vegetation within

certain aspects is not apparent (as it is within tower karst) and the overall distribution is

more dispersed. For slope, the vegetation coverage demonstrated similar distributions for

both tower karst and cockpit karst, particularly in the slope of 30-40 degrees.

44

Figure 15. The absence of vegetation on the south aspect of tower karst (Photo taken

during field survey of Guilin in Aug 2011)

3.5 Conclusion

Investigation of vegetation in tropical karst regions using GIS and RS has been

problematic because of shadows on the imagery. Previous studies either ignored the

shadows or primarily focused on non-shadow regions, leading to incomplete and

inaccurate results. In order to address this issue, this study proposed a comprehensive

approach of combing topographic correction, shadow retrieval and NDVI. The results

suggest that vegetation distribution varies between tower karst, and cockpit karst and

such differences correlate with the topographic characteristics of surface karst landforms.

In particular, the under-representation of vegetation in the south-southwest aspect of

tower karst was remarkable, and its overall distribution was less abundant and dispersed

than in cockpit karst. As a practical approach, the proposed method has considerable

45

potential for retrieving vegetation information when combined with topographic

correction and shadow restoration, which will provide solid bases for further geomorphic

and hydrologic analysis.

CHAPTER 4 HYDROLOGIC CONTROLS ON DEVELOPING TOWER KARST

AND COCKPIT KARST

4.1 Introduction

Hydrology plays an important role and defines the potential in developing karst

landforms because the chemical dissolution of limestones and other soluble rocks require

the interaction with water (Mangin, 1978). Hydrology is a major contributor to karst

development, and it provides a significant perspective to understanding the

characterization and evolution of karst landscapes (White, 2002). Karst hydrology is

complex, however, because the characteristic multiple permeability, involving pores,

fractures and conduits results in a wide range of permeabilities, rendering Darcy’s law

inapplicable in most cases (Ford et al. 1988). Therefore, soilless and water-scarce surface

conditions often result from high permeability and require a unique methodology in

hydrological analysis (LeGrand, 1973).

Jakucs (1977) defined the recharges to a karst zone from adjacent non-karstic areas as

allogenic and those from within the karst as autogenic. This duality of recharge is an

important component in understanding the hydrological controls over the development of

tropical karst. Other work (McDonald, 1975) described the role of allogenic drainage in

talus removal and the undermining of hillslope support in tower karst in Belize. Over

time, as the drainage continues to improve around the tower, channel incision and

46

migration result in less active hillslope erosion, making the steep tower slopes more

obscured at their bases. In his later work in Belize, McDonald (1979), stressed the

importance of rivers and recommended that this agency should be explored more widely

in interpreting karst landscape development, both within and beyond the tropics,

including the classical karst areas. McDonald argued that rivers are important agents of

geomorphic change, but are sometimes either not recognized or appear to be

underestimated in many karst landscapes. Where rivers are present, their potential role in

the geomorphic development of the karst landforms should be thoroughly evaluated,

including detailed field studies and the consideration of geological and sedimentological

evidence.

Footcaves at the bases of towers were identified by McDonald as an important cause of

steep flanks, eventual collapse and elimination of towers. They are associated with

surface and shallow ground waters active both in chemical processes and in “normal”

stream corrasion (McDonald and Twidale, 2011). Regardless of whether the towers stand

in isolation or in massifs, they are commonly riddled with caves and other voids

(Jennings, 1976). In humid tropical regions, where tower karst areas are traversed by

rivers flowing from non-karst areas, coarse detritus carried by the river will serve as the

tool of corrasion and undercut the bases of the towers (McDonald and Twidale, 2011).

McDonald (1975, 1976a, 1976b, 1979a,1979b; McDonald and Ley, 1985) reported that in

fenglin karst in Sarawak, Sulawesi, Belize, and Tabasco rivers in the monsoon/ wet

season flood the plains, making karst towers islands in ephemeral lakes. In addition,

McDonald also noted that sand and coarse fluvial debris in those areas had been

transported and deposited on valley floors and plains, provided that the area is either

47

close enough to the watertable for rivers and streams to remain at the surface, particularly

where they enter karst areas from a non-karst terrain. Such seasonal variation of allogenic

recharge was confirmed by Palmer (1975), who found this to be the case in a maze cave

bordering a non-karst catchment with high relief.

In addition to mechanical affects, allogenic recharge and internal runoff have powerful

undercutting capabilities because they are generally under-saturated with respect to

carbonate minerals and can initiate or accelerate the dissolution process in the carbonate

aquifer (White, 1988). As Miller (1987, p. **) stated in the case of Guatemala, “Fully-

integrated streams from non-carbonate highland appear to have been the primary factor in

developing the surface karst of the area; disaggregation of a fluvio-karst surface has

produced classic cockpit karst…”

Williams (1985) described the role of allogenic recharge in areas of pure, dense limestone

and a humid warm climate as leading to the formation of point recharge depressions.

Guilin seems to be an extreme example and, according to Yuan (1985) and Williams

(1980), tower karst in Guilin mainly occurs where surface runoff concurs with shallow

depth water-tables, whereas cockpit karst occurs where the water-table is below the

bottom of any enclosed basin. Using this comparison, Yuan (1985) concluded that tower

karst is not an icon of later phases of evolution and cockpit karst is not necessarily an

initial stage, and therefore a parallel rather than sequential model of development

deserves deep reconsideration.

Karst aquifers serve as intermediate layers that link surface runoff and groundwater,

allowing the surface water to go underground through infiltration (Ford et al. 1988). They

also represent valuable water resource for residents living in karst zones. Nonetheless,

48

since water contained in the aquifer may exist in both the vadose and the phreatic zones,

it may be difficult to access from the surface, and is always vulnerable to contamination

(Goldscheider and Ravbar, 2007).

One of the major challenges in karst groundwater modeling is the potential coexistence of

both Darcyian and non-Darcyain flow in the aquifer. Groundwater flow in pores, small

fractures and tight fissures is slow and may be approximated by Darcy’s law, but flow in

open fissures, large conduits and caves is turbulent (Kincaid, 2004), rendering Darcy’s

law inapplicable (Faulkner et al 2009). As a result, applying popular models such as

MODFLOW (Harbaugh, 2005) and MT3D will not guarantee correct results (Zheng and

Wang, 1999) that can be compared with field observation.

There are several approaches to this issue. One is to use observational data from wells or

pumping sites as references to generate hydrographs (Bonacci 1982 and 1995), although

other hydrologists argue that wells are independent of conduits, and thus cannot represent

the hydrology of karst aquifers (Ford and Williams 1989; Jeannin and Sauter 1998; Smart

1999). Another approach is to use lumped models that assume the aquifer is a “black box”

and predict its behavior by combining systematic inputs, outputs and transfer functions

(Martínez-Santos, 2010). Efforts to establish correlations between recharge, transfer and

discharge, however, neglect due consideration of the physics of groundwater flow (Zhang

et al., 1996 and Fleury et al., 2007). Another approach, the distributed model, is based on

the assumption that there is an equivalent porous medium in the aquifer, which requires

detailed information to render the hydrologic data with spatially coordinated elements

(Scanlon et al., 2003). Distributed models can also be elaborated by coupling Darcyian

and non-Darcyian flow (Stokes or Navier–Stokes systems) in the same model, but the

49

simulations rely heavily on detailed pre-investigation and expert knowledge of the karst

conduits and fissures within the aquifer (Doummar et al 2012). Therefore, such

applications are limited to small areas (Guo and Chen, 2006).

Solute transport mechanisms depend on the hydraulic behavior of the epikarst zone, the

flow out of which is a major factor that controls solute transport to the phreatic zone in

the aquifer. In particular, the hydraulic response of karst aquifers to storm events is

important to understanding solute transport mechanisms (Trček, 2008).

Solute transport in karst conduits is typically accomplished by rapid, turbulent flows

within a short period of time, while models of solute transport in surface water flow can

be determined empirically from quantitative ground water tracing studies (Field, 1997).

Few studies, have adopted a comprehensive approach to estimating the hydrological

control of tropical karst by combing groundwater modeling, solute transport modeling

and geographical analysis. Therefore, it could be advantageous to combine the

aforementioned techniques together to investigate the topic. The aim of this chapter is to

1) explore the water regime in the tower karst (fenglin) and cockpit karst (fengcong) of

Guilin, China; 2) combine both groundwater and solute transport models in the analysis,

and 3) estimate the bicarbonate concentration dynamics for both fenglin and fengcong.




50


of climatic, hydrological and geological conditions (Yuan, 1991, 2004; Zhu 1988).

Figure 16. Study area elevation map

Under the influence of monsoons from the Indian and Western Pacific Oceans, the study

area has pronounced dry and wet seasons, and 80 to 90% of total annual precipitation is

received from May to October (Zhao, 1986). The climate in Guilin is a subtropical

51

monsoon humid type (Liu 1991), with an annual average precipitation of 1873.6 mm

(Figure 17), annual average temperature of 18.8℃ (Figure 18), and annual average runoff

of 124 m3/s (Figure 19).

Figure 17. Annual precipitation of the study area

Figure 18. Annual temperature of the study area

Figure 19. Annual runoff of the study area

52

The soil sample data of sampling sites was acquired from Harmonized World Soil

Database (HWSD) viewer developed by the International Institute for Applied Systems

Analysis (IIASA). Bicarbonate concentration of wells and springs across the study area

was acquired from previous survey (Guilin Institute of Karst Geology, 1988)

4.3 Methodology

4.3.1 Watershed Delineation

There are several steps in the process of watershed delineation (Figure 20), including the

following:

1. Create a DEM without depressions

The first step in watershed delineation is to create a DEM (Digital Elevation Model)

without depressions. This can be achieved by filling sink cells and areas of internal

drainage in the original DEM data.

2. Calculate flow direction

Flow direction is calculated from each cell to its steepest downslope neighbors in the

format of a raster file.

3. Calculate flow accumulation

The calculation is conducted by creating accumulated flow to each cell during the process,

in which relevant streams, stream links and stream order were also generated. A threshold

of 5000 was applied to the stream network in order to create the streams in every pixel.

53

Figure 20. Methodology of delineating of watershed

4.3.2 Hydrological chart in the study area

Areal subsetting of

ASTER DEM

Flow Direction

No

DEM Filling

(Are there

any sinks?)

Yes

Depression-free

DEM

Flow Accumulation

Flow Length Stream Order

Stream Line

Stream Link

Watershed

Snap pour

54

The research regards fenglin and fengcong as subsystems in the study area, each of which

has its own hydrological inputs and outputs. The fengcong system has a thick vadose

zone underground and the allogenic flow from non karst areas can be ignored due to its

extremely small volume (see Figure 21). The fenglin system, on the other hand has

multiple inputs (including allogenic flow) and multiple outputs in the scheme (see Figure

22). It should be noted that the fenglin system involves longer response time than the

fengcong system because of its multiple inputs.

Figure 21. Hydrological chart of fengcong depression system (After Guilin karst geology,

1988)

Figure 22. Hydrological chart of fenglin plain system (After Guilin karst geology, 1988)

Comparison of annual rainfall and annual runoff indicated that the groundwater recharge

has significant contribution in the total water budget, so solute transport calculation in the

study will be primarily focused on the groundwater part (Figure 23).

55

Figure 23. Comparison of Annual Rainfall and Annual Runoff

4.3.3 Groundwater model and solute transport model

Richard’s equation was used for the groundwater flow in the one dimensional partially

saturated porous media

𝜕𝜃

𝜕𝑡=

𝜕

𝜕𝑧[𝐾 (

𝜕ℎ

𝜕𝑧+ 1)] − 𝑄

(4.1)

56

Where ℎ is the hydraulic head (𝐿), 𝜃 is the water content (𝐿3𝐿−3), 𝐾 is the hydraulic

conductivity (𝐿𝑇−1), 𝑡 is time (𝑇), 𝑧 is the spatial coordinate (𝐿), and 𝑄 is the sink-source

term (𝑇−1).

Due to the uneven nature of groundwater media, two different formulas (McDonald and

Harbaugh, 1988) were used in the two dimensional calculation. For a two dimensional

confined aquifer, the formula is

𝜕

𝜕𝑥[𝐾𝑥(ℎ − 𝑏)

𝜕ℎ

𝜕𝑥] +

𝜕

𝜕𝑦[𝐾𝑦(ℎ − 𝑏)

𝜕ℎ

𝜕𝑦] − 𝑊 = 𝑆

𝜕𝑦

𝜕𝑡

𝑇 = 𝑘(ℎ − 𝑏)

In the scenario of a two dimensional unconfined aquifer, the formula is

𝜕

𝜕𝑥(ℎ𝐾𝑥

𝜕ℎ

𝜕𝑥) +

𝜕

𝜕𝑦(ℎ𝐾𝑦

𝜕ℎ

𝜕𝑦) − 𝑊 = 𝑆

𝜕𝑦

𝜕𝑡

Where ℎ is the hydraulic head, 𝑘𝑥 and 𝑘𝑦 are the principal components of the hydraulic

conductivity tensor, 𝑇 is the transmissivity tensor, 𝑏 is the elevation of the bottom of the

aquifer, 𝑊 is the volumetric flux source or sinks of water, 𝑆 is the specific storage of

aquifer, 𝑡 is time.

The solute transport model for bicarbonate ion (Šimůnek and Suarez, 1994) was used in

the bicarbonate estimation.

𝜕(𝜃𝑐𝑘)

𝜕𝑡+ 𝜌

𝜕𝑐�̅�

𝜕𝑡+ 𝜌

𝜕�̂�𝑘

𝜕𝑡=

𝜕

𝜕𝑥𝑖(𝜃𝐷𝑖𝑗

𝜕𝑐𝑘

𝜕𝑥𝑗− 𝑞𝑖𝑐𝑘)

𝑘 = 1, 2, . . . , 𝑁𝑐

(4.2)

(4.3)

(4.4)

57

Where 𝑐𝑘 is the total dissolved concentration of the aqueous component 𝑘 (𝑀𝐿−3), 𝑐�̅� is

the total absorbed concentration of the aqueous component 𝑘 (𝑀𝑀−1), �̂�𝑘 is the total

precipitated concentration of the aqueous component 𝑘 (𝑀𝑀−1), 𝜌 is the bulk density of

the medium (𝑀𝐿−3), 𝐷𝑖𝑗 is the effective dispersion coefficient tensor (𝐿2𝑇−1), 𝑞𝑖 is the

volumetric flux (𝐿𝑇−1), and 𝑁𝑐 is the number of aqueous components.

4.3.4 Estimation of bicarbonate concentration

In order to estimate the bicarbonate concentration at sample sites from chapter 2

(geomorphic differentiation of fenglin and fengcong), water table gradients in the study

area were calculated from the ASTER DEM, then the flow path consisting of seepage

velocity and direction was calculated. The solute transport model was then applied to the

flow path for the estimation (Figure 24). Various observation results from previous

survey (Guilin karst geology, 1988) were used as initial injection point to predict the

bicarbonate content in the study area.

58

Figure 24. Bicarbonate estimation from solute transport model


According to the inquiry from the Harmonized World Soil Database (HWSD), the three

major soil types in the study area are Haplic Luvisols, Dystric Regosols and Cumulic

Anthrosols, with mixed rock outcrops (Figure 25). Since most of the sample area is

located in the Luvisols zone, Regosols and Anthrosols are not considered in the study.

Areal subsetting of

ASTER DEM

Calculate water table

gradient

Calculate Flow path

Application of Solute

transport model

Estimate the bicarbonate concentration at

selected sample sites

59

Figure 25. Soil types in the study area

Figure 26 demonstrates the pressure distribution at different depths of the sampling sites.

The vertical axis represents the depth of sample sites and the horizontal axis represents

the pressure head. The total time for the pressure simulation is 10 days with 5 intervals.

The initial blue line at zero seconds reaches the depth of 7 cm. As time progresses,

surface runoff penetrates to deeper depths and develops its own pressure profile, with

associated permeability.

60

Figure 26. Distribution of pressure (h) at different depth of sample sites

In figure 27, the simulation of bicarbonate transport follows the same time frame of ten

days (864,000 seconds). Runoff carrying bicarbonates infiltrates underground through

variably saturated porous media and the solute transport process is much slower than in

surface water. The ultimate concentration of bicarbonate after the elapse of 10 days

ranges between 0.2 and 0.3 kg/m3 (200-300 mg/l). The figure suggests that Darcyian flow

may also exist in the transport environment, but its roles in regulating groundwater is

constrained to limited fractions of the aquifer.

The result of bicarbonate concentration simulation is based on 100 samples and suggests

that, as might be expected, autogenic recharge from the carbonate area (light purple zone

of Figure 28) has a relatively higher concentration of bicarbonate, while the input from

the non-carbonate zone (rugged terrain with heavy shadows with blue and dark purple)

demonstrates lower concentrations of bicarbonate.

61

Figure 27.Solute transport of bicarbonate

Figure 28. Bicarbonate Concentration in the study area

62

4.5 Conclusion

Estimating the hydrological control on tropical karst is a big challenge due to the

heterogeneous nature of the porous aquifer. The study combined the ground water model

and solute transport model to estimate bicarbonate concentration. The results suggest that,

as would be anticipated, the permeabilities of karst aquifers in the study area vary in

depth, and that there are higher levels of bicarbonate ions within the karst area than in the

non-carbonate area.

CHAPTER 5 EDGE EFFECT

5.1 Introduction

Edge effects are widely acknowledged phenomena in ecology, typically occurring in

transitional sites where different habitats intersect. In particular, they occur as several

hundred meters-wide transition zones where adjacent but distinct ecosystems interact

with each other (Murcia, 1995) or where “natural” ecosystems abut adjacent disturbed or

developed land (Andren, 1995; Laurance, 1991; Laurance and Yensen, 1991; Kapos et al.,

1993; Lovejoy et al., 1986; Reed et al., 1996; Wilcove et al., 1986).

Many types of ecosystem edge effect dynamics have been identified, including changes

in the rate of leaf litter decomposition (Didham, 1998), changes in soil moisture, abrupt

changes in wind speed (Ranny et al., 1981), changes in light penetration and solar

radiation (Ranny et al., 1981), and changes in nutrient hydrological cycling (Saunders et

al., 1990).

63

Despite different types of ecosystem representation, edge effects are most common in

moderately to highly fragmented landscapes (Donovan et al., 1997; Hartley and Hunter,

1998; Thompson et al., 2002) and are less pronounced in large fragments where there is a

lower perimeter-area ratio (Helzer and Jelinski, 1999). Edge effects are greatest where

relatively small but different areas are juxtaposed (Keyser et al., 1998; Storch et al.,2005).

Although ecological edges are often recognized as borders between human-disturbed and

natural landscapes (Ewers and Didham, 2006; Lindenmayer and Fischer, 2006; Murcia,

1995), but they also occur where natural ecosystem transitions occur (Youngentob et al.,

2012), so they may have conceptual applicability both in karst to non-karst transitions

and within karst areas.

Karst in the Guilin area is a mixture of tower and cockpit karst styles, with transitions

between them (Zhu et al., 1988). Two groups of hypotheses present opposing views

regarding the evolution of tropical karst. Sequential models based upon Davis’s theory of

a Geographical Cycle (1899) have regarded tower karst and cockpit karst as different

stages of landscape evolution (Sweeting, 1958, 1990; Gerstenhauer, 1960; Song et al.,

1983; Williams, 1985; He, 1986), suggesting that tower karst has evolved from cockpit

karst, and that thus the former is an “older’ stage of the latter. By contrast, parallel

models have proposed that tower karst and cockpit karst have evolved simultaneously

without distinctive erosional stages (Verstappen, 1960, Balazs, 1968; Yuan 1981, 1986;

Williams, 1986; Yuan et al., 1990; Tan 1992; Sweeting, 1995 and Zhu et al., 1988).

Employing the ecological analogy, it is highly possible that there is an edge effect in

tropical karst where tower karst and cockpit coexist, particularly if the former represents

the edge effect or transition between karst and non-karst (Day and Huang, 2009). It is

64

hypothesized that tower karst necessarily occurs peripheral to or integrated with larger,

contiguous areas of cockpit karst, while the existence of cockpit karst does not

necessarily require the existence of tower in its proximity.

Geomorphological edge effects may both parallel and differ from biological edges in

terms of morphologies and processes. Initially, in this study the potential edge effects are

examined in terms of the geographic positioning of tower and cockpit karst landscapes,

with secondary attention to process variations. By contrast and comparison, biological

edge effects are concerned with the distribution, abundance and persistence of species, or

with biological process variationss (Ewers and Didham, 2006; Lindenmayer and Fischer,

2006; Murcia, 1995; Ries and Sisk, 2004).

This is the first study to try to identify the existence of edge effects in tropical karst, and

there has been no real consideration of how geomorphological/environmental variables

contribute to the formation of the edge. Defining the concept and testing the basic

hypothesis are thus of fundamental importance, and may ultimately help to explain the

evolution of tropical karst. This study of potential edge effects involving tower karst and

cockpit karst in the Guilin area is thus the first to define and attempt to identify this

phenomenon in tropical karst landscapes.

In order to examine the potential edge effects, this study will test the following

hypotheses: (1) Cockpit karst may develop independently on its own, and is not

necessarily associated with tower karst. (2) The development of tower karst requires

association with cockpit karst. (3) The coexistence of tower and cockpit landforms in a

given area reflects the similarity of their development conditions, but the two landforms

are the results of differing intensity of influential factors, such as NDVI, bicarbonate

65

concentrations, and surface drainage, exerted across the same open system over

heterogeneous temporal and spatial scales.


The study was conducted in Guilin, Guangxi, China, where there is a pronounced mixture

of tower karst and cockpit karst. The geographic location of tower karst and cockpit karst

are based on the classification results from chapter 2, corrected NDVIs are based on the

topographic correction and shadow restoration from chapter 3, and hydrological control

factors (bicarbonate ions) are from chapter 4.

5.3 Methodology

The nearest neighbor distances between tower karst (100 samples), cockpit karst (100

samples) and stream networks were calculated by using the proximity tools available via

the analysis tools in ArcGIS.

Geographic locations of tower karst (TK) and cockpit karst (CK) in both horizontal

(longitude) and vertical direction (latitude) were set as dependent variables, while the

nearest neighbor distance between tower karst, cockpit karst and stream networks,

corrected NDVI after topographic correction and shadow restoration, precipitation, and

the hydrological control factors (bicarbonate ions) were set as independent variables. One

hundred samples of fenglin and one hundred samples of fengcong were prepared for the

analysis, respectively. The data sets were introduced to IBM SPSS V22.0 software to

66

conduct a stepwise regression analysis. Prior to the regression analysis, correlation

between the independent variables was also checked.

After the stepwise regression analysis, the spatial autocorrelation analysis for fenglin and

fengcon in horizontal and vertical directions was obtained through the global Moran’s I

calculation.

5.4 Results and discussions

5.4.1 Stepwise regression results for fenglin (TK)

Tables 4 through 6 display the correlation between fenglin locations (TK_LAT) and the

influential factors in the vertical direction. NDVI, bicarbonate (BICAR), rainfall and

distant to fengcong (DIST2CK) have positive correlations with the vertical location of

fenglin. Results of stepwise regression (Table 5 and 6) show that the vertical distributions

of fenglin are determined by both distance to stream network and rainfall with an R

square of 0.807.

Tables 7 through 9 display the correlation between fenglin locations (TK_LON) and the

influential factors in the horizontal direction. Differing from the vertical direction result,

only bicarbonate (BICAR) has a positive correlation with the horizontal location of

fenglin, with other factors showing negative correlations. Results of stepwise regression

(Tables 8 and 9) show that the horizontal distribution of fenglin is determined by distance

to stream network, rainfall and concentration of bicarbonate, with an R square of 0.967.

67

Table 4. Correlation between tower karst latitude and other variables

Table 5. Model summary between tower karst latitude and other variables

Table 6. Model coefficients between tower karst latitude and other variables

68

Table 7. Correlation between tower karst longitude and other variables

Table 8. Model summary between Tower karst longitude and other variables

69

Table 9. Model coefficients between tower karst longitude and other variables

5.4.2 Stepwise regression results for fengcong (CK)

Tables 10 through 12 display the correlation between fengcong locations (CK_LAT) and

the influential factors in the vertical direction. NDVI, bicarbonate (BICAR) and rainfall

have positive correlations with the horizontal location of fengcong, with distance to

fenglin (DIST2TK) and distance to surface stream (DIST2HYDRO) showing negative

correlations. Results of stepwise regression (Tables 11 and 12) show that the vertical

distributions of fengcong are determined by bicarbonate, distance to stream network and

rainfall, with an R square of 0.970. It is negatively correlated with the distance to near

stream network (DIST2HYDRO), which suggests that the proximity of surface water

does not favor the development of fengcong in the vertical direction.

Tables 13 through 15 display the correlations between fengcong locations (CK_LON)

and the influential factors in the horizontal direction. Only the distance to fenglin

(DIST2TK) shows a positive correlation with the horizontal location of fengcong, with

NDVI, bicarbonate (BICAR), rainfall and distance to near surface stream

(DIST2HYDRO) showing negative correlations. Results of stepwise regression (Tables

70

14 and 15) show that the horizontal distributions of fengcong are determined by rainfall

and distance to stream network, with an R square of 0.922.

Table 10. Correlation between cockpit karst latitude and other variables

Table 11. Model summary between cockpit karst latitude and other variables

71

Table 12. Model coefficient between cockpit karst latitude and other variables

Table 13. Correlation between cockpit karst longitude and other variables

Table 14. Model summary between cockpit karst longitude and other variables

72

Table 15. Model coefficients between cockpit karst longitude and other variables

5.4.3 Spatial autocorrelation results for fenglin (TK)

The global Moran’s I for fenglin in the vertical direction (Figure 29) and horizontal

direction (Figure 30) show similar values of Moran’s index at around 0.186, suggesting

that they are somewhat random in overall spatial distribution across the study area.

Figure 29.Global Moran's I for Fenglin in vertical direction

73

Figure 30. Global Moran's I for Fenglin in horizontal direction

5.4.4 Spatial autocorrelation results for fengcong (CK)

The global Moran’s I for fengcong in the vertical direction (Figure 31) and horizontal

direction (Figure 32) show similar values of Moran’s index around 0.4 and a z-score

more than 0.6, suggesting that they are clustered pattern in overall spatial distribution

across the study area.

74

Figure 31. Global Moran's I for Fengcong in vertical direction

Figure 32. Global Moran's I for Fengcong in horizontal direction

75

The first hypothesis is thus supported by the regression analysis result (Tables 10 to 15),

because the distribution of fengcong is not strongly associated with the fenglin in both

(Distance to fenglin or DIST2TK) in both vertical and horizontal direction. In addition,

fengcong distribution is correlated with both rainfall (autogenic input) and distance to

hydro, suggesting that it can develop independently, and does not necessarily require an

association with fenglin.

Concerning the second hypothesis, the study supports the hypothesis that the

development of fenglin does require an association with fengcong, particularly in the

horizontal direction. This is because the bicarbonate (allogenic input) is generally

injected in the bordering fengcong area, then it will pass through the groundwater

network and discharge from the outlet or spring in the fenglin area (Ford and Williams,

1989).

The study also supports the third hypothesis. The regression analysis and spatial

autocorrelation analysis, show that fenglin and fengcong may coexist in the same area,

but that their respective development is associated with variable intensities of different

input factors, such as NDVI, bicarbonate concentrations, rainfall and surface drainage.

The vegetation factor (NDVI) is not a determining factor in either fenglin or fengcong

distribution probably because it is outweighed by hydrological factors and bicarbonate

distributions.

5.5 Conclusion

The study investigates the edge effects in tropical karst areas using a case study of Guilin

in a fragmented tower-cockpit karst landscape. The results show that the edge effect does

exist in the Guilin area, with variable intensities in different directions (vertical and

76

horizontal). In addition, the study also demonstrates that the fenglin distribution exhibits

a random pattern while the fengcong is more clustered in the study area.

CHAPTER 6 CONCLUSIONS

6.1 Summary

Tower karst (fenglin) and cockpit karst (fengcong) are two important representative

styles in tropical karst regions. Previous studies proposed sequential and parallel models

of evolution, but they provide insufficient evidence to illuminate the spatial and temporal

relationships between the two landforms. This unclear interpretation of tower-cockpit

relationships both obscures understanding of the process-form dynamics, and also

confuses the morphological definitions. Prior technological limitations, as well as the

fragmental nature of the karst landscape, limited incorporation of geologic and

hydrologic data into broad geospatial frameworks using GIS and RS techniques, and

most of this data is scattered. Postulated edge effects have not previously been

investigated, and little is known about how different environmental variables

(geomorphology, vegetation, hydrological control, precipitation) contribute to the

dynamics and morphology edges. To address these issues, this research has combined

geographic, geologic and hydrologic data using GIS and RS technologies to generate

solid and reliable quantitative evidence of the edge effect. Specifically, there were be four

purposes of the study. The first was to develop an effective method of differentiating

fenglin and fengcong. The second was to extract the vegetation information without

ignoring the shadows on the satellite imagery, and investigate the correlation between the

karst topography and its vegetation. The third was to combine the regional hydrology and

solute transport models to estimate geochemical controls over fenglin and fengcong. The

77

fourth, perhaps the most important objective, was to test the edge effect hypothesis using

the results from the aforementioned three components.

The results of the dissertation lead to several significant conclusions. First, DEM data are

very useful for extracting profiles of complicated surface landforms with shadows on the

satellite imagery. Second, the vegetation distribution varies within and between tower

karst and cockpit karst, and such differences correlate with their topographic

characteristics. The under-representation of vegetation in the south-southwest aspect of

tower karst was remarkable, and its overall distribution is less abundant and dispersed

than in cockpit karst. Third, testing supported the edge effect hypothesis, showing

variable intensity and extension in different directions. Additionally, the study also shows

that the fenglin is distributed in a random pattern, while the fengcong is clustered within

the study area.

6.2 Contribution

The first contribution of this study is highlighted by developing a method to classify

tropical karst effectively in the presence of shadows that would otherwise hinder

traditional approaches from classifying the complicated landforms. Second, the study

suggests non-random variance of vegetation vitality in certain aspects of fenglin that

could explain its geomorphic difference from fengcong. Third, the study developed a

method to combine the groundwater and solute transport models to estimate bicarbonate

concentration distribution. Finally, for the first time, the study supported the edge effect

hypothesis through a systematic and quantitative approach, which may lead to new

insights via similar studies in the future.

78

6.3 Future research

Future research might usefully attempt to employ similar methodologies to other regions

with similar climatic and geologic conditions that promote the development of the tower

and cockpit styles of tropical karst. Central America (Belize, Guatemala), the Caribbean

(Jamaica, Puerto Rico, Hispaniola and Cuba) and Southeastern Asia (Indonesia, Malaysia,

Vietnam) constitute appropriate study areas where comparable results might be produced

through similar analysis. It would be particularly valuable to ascertain whether the

demonstrated cockpit-tower edge effect is universal or is constrained within certain

geographic locations.

79

REFERENCES

Andrén, H., 1995. Effects of landscape composition on predation rates at habitat edges.

Mosaic landscapes and ecological processes, ed. Hansson, L., Fahrig, L., and Merriam,

G. Springer Netherlands, Chapman and Hall, London, pp 225-255

Asner, G.P., & Heidebrecht, K.B., 2003. Imaging spectroscopy for desertification studies:

comparing AVIRIS and EO-1 Hyperion in Argentina drylands. IEEE Transactions on

Geoscience and Remote Sensing, 41, 1283-1296.

Asner, G.P., Knapp, D.E., Cooper, A.N., Bustamante, M.M.C., Olander, L.P., &

Heidebrecht, K.B., 2005. Ecosystem structure throughout the Brazilian amazon from

Landsat observations and automated spectral unmixing, Earth Interactions, 9 (1)

Balazs, D., 1973. Relief types of tropical karst areas. Proceedings of the Symposium on

Karst Morphogenesis. Proceedings of International Geographical Union Regional

Conference (Hungary), 16–32.

Bolch, T., Kamp, U., Olsenholler, J., 2005. Using ASTER and SRTM DEMs for studying

geomorphology and glaciation in high mountain areas. New Strategies for European

Remote Sensing, ed. Oluic, M. Rotterdam: Millpress, 119–128.

Bonacci, O., 1982. Specific hydrometry of karst regions. Advances in Hydrometry, IAHS

Publ. 134, 321-333.

Bonacci, O., 1995. Ground water behaviour in karst: example of the Ombla Spring

(Croatia). Journal of Hydrology, 165(1), 113-134.

80

Brook, G.A. and Hanson, M., 1991, Double Fourier series analysis of cockpit and doline

karst near Browns Town, Jamaica. Physical Geography, 12(1), 37-54

Bubenzer, O. and Bolten, A., 2008. The use of new elevation data (SRTM/ASTER) for

the detection and morphometric quantification of Pleistocene megadunes (draa) in the

eastern Sahara and the southern Namib. Geomorphology, 102, 221–231.

Chen, Y., Wen, D., Jing, L., and Shi, P., 2007. Shadow information recovery in urban

areas from very high resolution satellite imagery. International Journal of Remote

Sensing, 28(15): 3249-3254.

Chenoweth, M.S. and Day, M.J, 2001. Developing a GIS for the Jamaican Cockpit

Country. Geotechnical and environmental applications of karst geology and hydrology,

ed. Beck B.F. and Herring J.G. Swets & Zeitlinger Publishers, Lisse, Netherlands, 67–71

Dare, P.M., 2005. Shadow analysis in high-resolution satellite imagery of urban areas.

Photogrammetric Engineering and Remote Sensing, 71,169–177

Day, M. J., 1978. Morphology and distribution of residual limestone hill in the karst of

northern Puerto Rico. Geological Society of America Bulletin, 89(3), 426-432

Day, M.J. 1979. Surface roughness as a discriminator of tropical karst styles. Zeitschrift

fur Geomorphologie, Supplementbande, 32, 1-8.

Day, M.J., 1981. Towards numerical categorization of tropical karst terrains. Proceedings

of 8th International Congress of Speleology, 1, 330-332.

Day, M. J., 1982. The influence of some material properties on the development of

tropical karst terrains. Transactions of British Cave Research Association, 9 (1), 27-37

81

Day, M.J., 2002. The role of valley systems in the evolution of tropical karstlands. In:

Evolution of Karst: From Prekarst to Cessation, ed. F. Gabrovsek. Zalozba ZRC,

Ljublana, 235-241.

Day, M.J., 2004a. Cone karst, in Encyclopedia of Caves and Karst Science, ed. Gunn, J.

New York: Fitzroy Dearborn, 241-243.

Day, M. J., 2004b. Morphometry of Karst, in Encyclopedia of Caves and Karst Science,

ed. Gunn, J. New York: Fitzroy Dearborn, 526-527.

Day, M.J. and Chenoweth, M.S., 2004. The Karstlands of Trinidad and Tobago, their

land use and conservation. The Geographical Journal, 170 (3), 256-266.

Day, M.J. and Chenoweth, M.S., 2013a. Cockpit Country Cone Karst, Jamaica,

Encyclopedia of Caves and Karst Science, ed. Gunn, J. New York: Fitzroy Dearborn,

233-235.

Day, M.J. and Chenoweth, M.S., 2013b. Surface roughness of karst landscapes, Treatise

on Geomorphology, Volume 6, Karst Geomorphology, ed. Frumkin, A, New York:

Academic Press: 157-163.

Day, M.J. and Tang, T., 2004. Tower karst. Encyclopedia of Caves and Karst Science,

ed. Gunn, J. New York: Fitzroy Dearborn, 734-736.

Day, M.J. and Huang, W., 2009. Reflections on fengcong and fenglin. Cave and Karst

Science, 36(2), 49-51.

Didham, R.K., 1998. Altered leaf-litter decomposition rates in tropical forest fragments.

Oecologia, 116(3), 397-406.

Donovan, T.M., Jones, P.W., Annand, E.M., and Thompson III, F.R., 1997.Variation in

82

local-scale edge effects: mechanisms and landscape context. Ecology, 78(7), 2064–

2075.

Doummar, J., Sauter, M., and Geyer, T., 2012. Simulation of flow processes in a large

scale karst system with an integrated catchment model (Mike She)–Identification of

relevant parameters influencing spring discharge. Journal of Hydrology, 426, 112-123.

Ewers, R.M., and Didham, R.K., 2006. Confounding factors in the detection of species

responses to habitat fragmentation. Biological Reviews, 81(1), 117–142.

Faulkner, J., Hu, B. X., Kish, S., and Hua, F., 2009. Laboratory analog and numerical

study of groundwater flow and solute transport in a karst aquifer with conduit and matrix

domains. Journal of contaminant hydrology, 110(1), 34-44.

Field, M.S., 1997. Risk assessment methodology for karst aquifers:(2) solute-transport

modeling. Environmental Monitoring and Assessment, 47(1), 23-37.

Fleurant, C., Tucker, G.E. and Viles, H.A. 2007. Modele d’evolution de paysages,

application au karstes en cockpit de Jamaique. Karstologia, 49: 33-42.

Fleurant, C., Tucker, G.E. and Viles, H.A. 2008a. Modelling cockpit karst landforms. in:

Landscape Evolution: Denudation, Climate and Tectonics over Different Time and Space

Scales, ed. Gallagher, K., Jones, S.J. and Wainwright, J. London: Geological Society

Special Publication 296: 47-62

Fleurant, C., Tucker, G.E. and Viles, H.A. 2008b. A model of cockpit karst landscape,

Jamaica. Géomorphologie: relief, processus, environnement, 1, 3-14.

http://geomorphologie.reviews.org/index5653.html

83

Fleury, P., Plagnes, V., and Bakalowicz, M., 2007. Modelling of the functioning of karst

aquifers with a reservoir model: Application to Fontaine de Vaucluse (South of France).

Journal of hydrology, 345(1), 38-49.

Ford, D. C., Palmer, A. N., and White, W. B., 1988. Landform development; karst.

Geological Society of America. 2, 401-412

Ford, D.C., and Williams, P.W., 1989. Karst geomorphology and hydrology. London:

Unwin Hyman, 601pp

Ford, D.C. and Williams, P.W., 2007. Karst Hydrologeology and Geomorphology,

Chichester; Wiley, 562pp.

Gao, Y., Zhang, W., 2009. LULC classification and Topographic Correction of Landsat-7

ETM + Imagery in the Yangjia RivernWatershed: The influence of DEM Resolution

Sensors. Sensor, 9(3): 1980–1995.

Gellert, J.F. 1962. Der Tropenkarst in Sub-china im Rahmen der Gebirgsformung des

landes. Verhandlungen des Deutschen Geographentages, 33, 376-84.

Giles, P.T., 2001. Remote sensing and cast shadows in mountainous terrain. Journal of

the American Society for Photogrammetry and Remote Sensing, 67(7), 833-839

Goldscheider, N., and Ravbar, N. 2007. Proposed methodology of vulnerability and

contamination risk mapping for the protection of karst aquifers in Slovenia. Acta

carsologica, 36(3), 397-411

Guilin Karst and Geological Structure, 1988, Chongqing Pressing House, 130 pp

84

Guo, L., and Chen, Z., 2006. The Puzzles and Solutions for Quantitative Assessment of

Water Resource in Karst Underground River System. Carsologica Sinica, 25(1), 1-5 (in

Chinese)

Haboudane, D., Miller, J.R. and Pattey, E., 2004. Hyperspectral Vegetation Indices and

Novel Algorithms for Predicting Green LAI of Crop Canopies: Modeling and Validation

in the Context of Precision Agriculture. Remote Sensing of Environment, 90: 337–352.

Harbaugh, A. W., 2005. MODFLOW-2005, the US Geological Survey modular ground-

water model: The ground-water flow process (pp. 6-A16). US Department of the Interior,

US Geological Survey.

Harden, C.P., 2002. Hillslope runoff, soil detachment, and soil organic content following

reforestation in the Copper Basin, Tennessee, USA. Australian Geographical Studies 40

(2), 130–142.

Hartley, M.J., and Hunter, M.L., 1998. A meta-analysis of forest cover, edge effects, and

artificial nest predation rates. Conservation Biology, 12(2), 465–469.

Helzer, C.J., and Jelinski, D.E., 1999. The relative importance of patch area and

perimeter area ratio to grassland breeding birds. Ecological applications, 9(4), 1448–

1458.

Hodgson, M. E., Jensen, J. R., Tullis, J. A., Riordan, K. D., & Archer, C. M., 2003.

Synergistic use of Lidar and color aerial photography for mapping urban parcel

imperviousness. Photogrammetric Engineering and Remote Sensing, 69, 973–980.

85

Huang, J., Yan, Q., Wang, M., Zhou, W., Gua, C., Shi, J., Pei, J., Huang, J. and Li,

Q.,1988. Study on karst water resource evaluations in Guilin and its methodology.

Chongqing Publishing House. 134 pp. (in Chinese)

Huang, W., Deng, C., and Day, M.J., 2014. Differentiating tower karst (fenglin) and

cockpit karst (fengcong) using DEM contour, slope, and centroid. Environmental Earth

Sciences. 72(2): 407-416

Huang, Z., Turner, B. J., Dury, S. J., Wallis, I. R., & Foley, W. J. 2004. Estimating

foliage nitrogen concentration from HYMAP data using continuum removal analysis.

Remote Sensing of Environment, 93(1), 18-29.

Hutchinson, M.F. and Gallant, J.C., 2000. Digital elevation model and representation of

terrain shape. Terrain Analysis: Principle and Application, ed. Gallant, J.C and Wilson,

J.P. New York: Wiley: 29-50.

Jakucs, L., 1977. Morphogenetics of karst regions: Variants of karst evolution. Budapest:

Akademiai Kiado, p284

Jeannin, P.Y., and Sauter, M., 1998. Analysis of karst hydrodynamic behaviour using

global approaches: a review. Bulletin Centre d'Hydrogéologie, Université de Neuchatel,

16, 31-48.

Jensen, J., 2007. Introductory Digital Image Processing. Beijing: Science Press and

Pearson Education Asia Limited, China, 127–173, 220–221.

Jones, A. R., Settle, J. J., and Wyatt, B. K., 1988. Use of digital terrain data in the

interpretation of SPOT-1 HRV multispectral imagery. Remote Sensing, 9(4), 669-682.

86

Kapos, V., Ganade, G., Matsui, E., and Victoria, R.L., 1993. Carbon 13 isotope as an

indicator of edge effects in tropical rainforest reserves. Journal of Ecology, 81, 425-432.

Keesstra, S.D., van Dam, O., Verstraeten, G., van Huissteden, J., 2009. Changing

sediment dynamics due to natural reforestation in the Dragonja catchment, SW Slovenia.

Catena, 78 (1), 60–71.

Keyser, A. J., Hill, G. E., and Soehren, E.C., 1998. Effects of Forest Fragment Size, Nest

Density, and Proximity to Edge on the Risk of Predation to Ground‐Nesting Passerine

Birds. Conservation Biology, 12(5), 986-994.

Kim, M.K., and Daigle, J.J., 2011. Detecting Vegetation Cover Change on the Summit of

Cadillac Mountain Using Multi-Temporal Remote Sensing Datasets: 1979, 2001, and

2007. Environmental Monitoring and Assessment, 180, 63–75.

Kincaid, T., 2004. Exploring the secrets of Wakulla Springs. In open seminar,

Tallahassee.

Kohavi, R and Provost, F., 1998. Glossary of Terms. In: Editorial for the Special Issue on

Applications of Machine Learning and the Knowledge Discovery Process, Machine

Learning, 30, 271-274

Kokaly, R. F., 2001. Investigating a physical basis for spectroscopic estimates of leaf

nitrogen concentration. Remote Sensing of Environment, 75, 153-161.

Laurance, W.F. 1991. Edge effects in tropical forest fragments: application of a model for

the design of nature reserves. Biological Conservation, 57(2), 205-219.

Laurance, W.F. and Yensen, E., 1991. Predicting the impacts of edge effects in

fragmented habitats. Biological Conservation, 55(1), 77-92.

87

LeGrand, H. E., and Stringfield, V. T., 1973. Karst hydrology—a review. Journal of

Hydrology, 20(2), 97-120.

Liang, F and Xu, B., 2013. Discrimination of tower-, cockpit-, and non-karst landforms in

Guilin, Southern China, based on morphometric characteristics, Geomorphology, 204,

42-48

Lindenmayer, D.B., and Fischer, J., 2006. Habitat Fragmentation and Landscape Change:

an Ecological and Conservation Synthesis. Island Press, Washington, DC.

Liu, W., and Yamazaki, F., 2012. Object-based shadow extraction and correction of high-

resolution optical satellite images. IEEE Journal of Selected Topics in Applied Earth

Observation and Remote Sesning, 5(4), 1296–1302.

Liu, Z., 1991. IGCP 299 "Geology, Climate, Hydrology and Karst Formation" Guidebook

for Field Excursions of International Symposium and Field Seminar on Karst of Inner

Plate Region with Monsoon Climate. P1-18. The Institute of Karst Geology

Lovejoy, T.E., Bierregaard, Jr. R.O., Rylands, A.B., Malcolm, J.R., Quintela, C.E.,

Harper, L. H., Brown, K.S., Powell, A.H., Powell, G.V.N., Schubert, H.O.R., and Hays,

M.B., 1986. Edge and other effects of isolation on Amazon forest fragments.

Conservation Biology: The Science and Scarcity of Diversity, ed. Soule, M. Sunderland,

Massachusetts, pp7-12

Lyew-Ayee, P. 2004. Digital Topographic Analysis of Cockpit Karst: A

Morphogeological Study of the Cockpit Country Region, Jamaica, PhD thesis, University

of Oxford.

88

Lyew-Ayee, P., Viles, H.A. and Tucker, G.E. 2007. The use of GIS-based digital

morphometric techniques in the study of cockpit karst. Earth Surface Processes and

Landforms, 32(2), 165-179.

Mangin, A., 1978. Le karst, entité physique, abordé par l’étude du système karstique.

(Karst as a physical unit, from the study of the karst system) Le karst, colloque de Tarbes,

17-18

Martínez-Santos, P., and Andreu, J. M., 2010. Lumped and distributed approaches to

model natural recharge in semiarid karst aquifers. Journal of hydrology, 388(3), 389-398.

Marston, R.A., Bravard, J.-P., Green, T., 2003. Impacts of reforestation and gravel

mining on the Malnant River, Haute Savoie, French Alps. Geomorphology 55 (1–4), 65–

74.

Marston, R.A., 2010. Geomorphology and vegetation on hillslopes: Interactions,

dependencies, and feedback loops. Geomorphology, 116, 206–217

Mather, P.M., 2004. Computer Processing of Remotely-Sensed Images. London: John

Wiley & Sons Ltd., 81 and 136.

McDonald, M.G., and Harbaugh, A.W., 1988. A modular three-dimensional finite-

difference groundwater flow model. Techniques of water resources investigations, 06-A

1. US Geophys Soc.

McDonald, R.C. 1975. Observations on hillslope erosion in tower karst topography of

Belize. Geological Society of America Bulletin, 86(2), 255–256.

89

McDonald, R.C. 1976a. Limestone morphology in South Sulawesi, Indonesia. Zeitschrift

für Geomorphologie, Supp, 26, 79-91.

McDonald, R.C. 1976b. Hillslope base depressions in tower karst topography of Belize.

Zeitschrift für Geomorphologie, Supp, 26, 98-103.

McDonald, R.C. 1979a. Tower karst geomorphology in Belize. Zeitschrift für

Geomorphologie Supp, 32, 35-45

McDonald, R.C. 1979b. Tower Karst Geomorphology, with Special Reference to Belize,

Indonesia, and Malaysia. Doctoral Dissertation, University of Oxford.

McDonald, R.C. 2002. Rivers in karst geomorphology, in Evolution of Karst: From

Prekarst to Cessation, ed. Gabrovsek, F. Ljubljana: Zalozba SRC: 267-273

McDonald, R.C and Ley, R G. 1985. Tower karst geomorphology in northern Borneo.

Zeitschrift für Geomorphologie, 29, 483–495.

McDonald, R.C and Twidale C.R., 2011, On the Origin and Significance of Basal

Notches or Footcaves in Karst Terrains, Physical Geography, 32(3), 195-216

Miller, T., 1987. Fluvial and collapse influence on cockpit karst of Belize and eastern

Guatemala. Karst Hydrogeology: Engineering and Environmental Applications, ed. B.F.

Beck. Rotterdam: AA Balkema, 53-63

Murcia, C., 1995. Edge effects in fragmented forests: implications for conservation.

Trends in Ecology and Evolution, 10 (2), 58–62.

90

Mylroie, J.E. and Mylroie, J.R., 2009. Caves as sea level and uplift indicators, Kangaroo

Island, South Australia. Journal of Cave and Karst Studies, 71(1), 32-47.

Nakajima, T., Tao, G., and Yasuoka, Y., 2002. Simulated recovery of information in

shadow areas on IKONOS image by combing ALS data. Proceedings of Asian

Conference on Remote Sensing. Retrieved at http://www. gisdevelopment.

com/aars/acrs/2002/vhr/214. pdf.

Palmer, A. N. 1975. The origin of maze caves. National Speleological Society Bulletin,

37(3), 56-76.

Parise, M., J. D. Wales, and F. Gutierrez. 2009. Current Perspectives on the

Environmental Impacts and Hazards in Karst. Environmental Geology, 58, 235–237.

Purkis, S.J., Rowlands, G.P., Riegl, B.M., and Renaud, P.G., 2010, The paradox of

tropical karst morphology in the coral reefs of the arid Middle East. Geology, 38, 227–

230.

Ranney, J.W., Bruner, M.C., and Levenson, J.B., 1981. The importance of edge in the

structure and dynamics of forest islands. Forest Island Dynamics in Man-Dominated

Landscapes, ed. Burgess a, R.L. and Sharpe, D.M. Springer Verlag, New York. pp 67-95

Reed, R. A., Johnson-Barnard, J., and Baker, W.L., 1996. Fragmentation of a forested

Rocky Mountain landscape, 1950–1993. Biological Conservation, 75(3), 267-277.

Reinhardt, L., Jerolmack, D., Cardinale, B.J., Vanacker, V., Wright, J., 2010. Dynamic

interactions of life and its landscape: feedbacks at the interface of geomorphology and

ecology. Earth Surface Processes and Landforms, 35, 78–101.

91

Renschler, C.S., Doyle, M.W., Thoms, M., 2007. Geomorphology and ecosystems:

challenges and keys for success in bridging disciplines. Geomorphology, 89 (1–2),1–8.

Riaño, D., Chuvieco, E., Salas, J., and Aguado, I., 2003. Assessment of different

topographic corrections in Landsat-TM data for mapping vegetation types. IEEE

Transactions on Geoscience and Remote Sensing, 41(5), 1056-1061.

Ries, L., and Sisk, T.D., 2004. A predictive model of edge effects. Ecology, 85(11),

2917–2926.

Roberts, D. A.; Smith, M. O.; Adams, J. B., 1993. Green Vegetation, Nonphotosynthetic

Vegetation, and Soils in AVIRIS Data. Remote Sensing of Environment, 44 (2-3), 255-

269

Ru, J. et al.,1988. The comprehensive study of Lijiang River Watershed Management,

Guangxi Normal University Publishing House, 59 pp (In Chinese)

Salvador, E., Cavallaro, A and Ebrahimi,T., 2001. Shadow identification and

classification using invariant color models. Proceedings of IEEE International

Conference of Acoustics Speech Signal Process. vol.3, 1545–1548.

Sarabandi, P., Yamazaki, F., Matsuoka, M., and Kiremidjian, A., 2004. Shadow detection

and radiometric restoration in satellite high resolution images. Proceedings of IGARSS-

2004, September 2004, Anchorage, Alaska (New York: IEEE), CDROM.

Saunders, D.A., Hobbs, R.J., and Margules, C.R. 1990. Biological consequences of

ecosystem fragmentation: a review. Conservation Biology, 5(1), 18-32.

92

Scanlon, B.R., Mace, R.E., Barrett, M.E., and Smith, B., 2003. Can we simulate regional

groundwater flow in a karst system using equivalent porous media models? Case study,

Barton Springs Edwards aquifer, USA. Journal of Hydrology, 276(1), 137-158.

Shackelford, A. K., and Davis, C. H., 2003. A hierarchical fuzzy classification approach

for high-resolution multispectral data over urban areas. IEEE Transactions on

Geoscience and Remote Sensing, 41(9), 1920-1932.

Shahtahmassebi, A., Yang, N., Wang, K., Moore, N., and Shen, Z.,2013. Review of

shadow detection and de-shadowing methods in remote sensing. Chinese Geographical

Science, 23(4), 403-420.

Šimůnek,J., and Suarez, D.L., 1994. Two dimensional transport model for variably

saturated porous media with major ion chemistry. Water Resources Research, 30(4),

1115-1133.

Smart, C.C., 1999. Subsidiary conduit systems: a hiatus in aquifer monitoring and

modelling. Proceedings of Karst modelling Symposium, ed. Palmer, A.N., Palmer, M.V.,

and Sasowsky, I.D. Charlottesville, February 24–27, 1999. Karst Water Institute, Spec

Publ 5, 146–157

Smart, P., Waltham, T., Yang, M. and Zhang, Y., 1986. Karst geomorphology of western

Guizhou, China. Cave Science 13: 89-103.

Sweeting, M.M., 1972. Karst Landforms. London and New York: Macmillan, 362pp.

Tang, T. and Day, M.J. 2000. Field survey and analysis of hillslopes on tower karst in

Guilin, southern China. Earth Surface Processes and Landforms, 25: 1221-1235.

93

Teillet, P.M., Guindon, B., and Goodenough, D.G., 1982. On the slope-aspect correction

of multispectral scanner data, Canadian Journal of Remote Sensing, 8(2), 84-106

Trimble, 2011. eCognition Developer 8.64.1-User Guide, Munchen: Trimble Germany

GmbH., 242pp.

Trček, B., 2008. Flow and solute transport monitoring in the karst aquifer in SW Slovenia.

Environmental geology, 55(2), 269-276.

Troester, J.W., 1992. The northern karst belt of Puerto Rico: A humid tropical karst,

Hydrogeology of selected karst regions of the world: International Association of

Hydrogeologists, International Contributions to Hydrogeology, 13, 475-486.

Tsai, V. J., 2006. A comparative study on shadow compensation of color aerial images in

invariant color models. IEEE Transactions on Geoscience and Remote Sensing, 44(6),

1661-1671.

Tucker, C.J., 1979. Red and photographic infrared linear combinations for monitoring

vegetation. Remote Sensing of Environment, 8, 127-150.

Waltham, T. 2008. Fengcong, fenglin, cone karst and tower karst. Cave and Karst

Science, 35(3), 77–88.

Ustin, S.L., Roberts, D.A., Gamon, J.A., Asner, G.P., & Green, R.O., 2004. Using

imaging spectroscopy to study ecosystem processes and properties. Bioscience, 54, 523-

534.

Wang, S., Liu, Q., and Zhang, D., 2004. Karst Rocky Desertification in Southwestern

China: Geomorphology, Landuse, Impact and Rehabilitation. Land Degradation and

Development, 15, 115–121.

94

Wang, S., and Li, Y., 2007. Problems and Development Trends about Researches on

Karst Rocky Desertification. Advance in Earth Sciences, 6, 573–582.

Waltham, T., 2008. Fengcong, fenglin, cone karst and tower karst. Cave and Karst

Science, 35(3), 77–88.

Weng, Q., 2012. Remote sensing of impervious surfaces in the urban areas: Requirements,

methods, and trends. Remote Sensing of Environment, 117, 34–49

White, W.B., 1988. Geomorphology and hydrology of karst terrains. New York: Oxford

university press. pp 464

White, W.B., 2002. Karst hydrology: recent developments and open questions.

Engineering geology, 65(2), 85-105.

Williams, P.W., 1972. Morphometric analysis of polygonal karst in New Guinea.

Geological Society of America Bulletin, 83(3), 761-796.

Williams, P.W., 1980, Hydrological control and the development of cockpit and tower

karst, Proceedings of IAH 21st Congress of Karst Hydrogeology and karst environment

protection, Guilin, China, 281-290

Williams, P. W., 1985. Subcutaneous hydrology and the development of doline and

cockpit karst. Zeitschrift für Geomorphologie, 29(4), 463-482.

Williams, P.W., 1987. Geomorphic inheritance and the development of tower karst.

Earth Surface Processes and Landforms, 12(5): 453-465.

Yang, X., Skidmore, A. K., Melick, D., Zhou, Z., and Xu, J., 2007. Towards an

efficacious method of using Landsat TM imagery to map forest in complex mountain

terrain in Northwest Yunnan, China. Tropical Ecology, 48(2), 227.

95

Yao, J. and Zhang, Z., 2006. Hierarchical shadow detection for color aerial images,

Computer Vision and Image Understanding, 102, 60–69

Yesilnacar, E., and Süzen, M.L., 2006. A land‐ cover classification for landslide

susceptibility mapping by using feature components. International Journal of Remote

Sensing, 27(2): 253-275.

Yuan, D., 1984. About Fengcong karst. Geology of Guangxi 1, 79–84 (in Chinese).

Yuan, D., 1985. On the heterogeneity of karst water. Karst water Resource. Proceedings

of the Ankara-Antalya Symposium, IAHS Publ, 161, 281-292

Yuan, D., 1991. Karst of China. Beijing: Geological Publishing House, 224pp.

Yuan, D., 1992. Karst in Southwest China and its comparison with karst in north China.

Quaternary Sciences of China. 4, 352-361 (in Chinese)

Yuan D., 2004. Yangshuo karst, China, in Encyclopedia of Caves and Karst Science, ed.

Gunn, J. Fitzroy Dearborn: New York, 781-783.

Yuan, D., and Zhang, C., 2008. Karst Dynamics Theory in China and Its Practice. Acta

Geoscientica Sinica, 29: 355–365.

Yue, Y., Zhang, B., Wang, K., Liu, B., Li, R., Jiao, Q., Yang, Q., Zhang, M., 2010.

Spectral indices for estimating ecological indicators of karst rocky desertification,

International Journal of Remote Sensing, 31(8), 2115-2122.

Yue, Y., Wang, K., Liu, B., Li, R., Zhang, B., Chen, H., and Zhang, M., 2013.

Development of new remote sensing methods for mapping green vegetation and exposed

96

bedrock fractions within heterogeneous landscapes. International Journal of Remote

Sensing, 34:14, 5136-5153.

Zeng, Z., 1982. Characters of the karst topography in southern China. Carsologica

Sinica ,1, 27–31 (in Chinese).

Zhan, Q., Shi, W., and Xiao, Y., 2005. Quantitative analysis of shadow effects in high-

resolution images of urban areas. International Archives of Photogrammetry and Remote

Sensing, 36(8/W27). Available at: http://www.isprs.org/proceedings/XXXVI/8-

W27/zhan.pdf

Zhang, Y., Bai, E., Libra, R., Rowden, R., and Liu, H., 1996. Simulation of spring

discharge from a limestone aquifer in Iowa, USA. Hydrogeology Journal, 4(4), 41-54.

Zhang, Z., De Wulf, R. R., Van Coillie, F.M., Verbeke, L.P., De Clercq, E.M., and Ou X.,

2011. Influence of different topographic correction strategies on mountain vegetation

classification accuracy in the Lancang Watershed, China, Journal of Applied Remote

Sensing, 5(1), 053512-053521

Zhao, S., 1986, Physical Geography of China. New York and London: Wiley, 209pp

Zhou, Y., Chen, J., Guo, Q., Cao, R., and Zhu, X., 2014. Restoration of Information

Obscured by Mountainous Shadows Through Landsat TM/ETM+ Images Without the

Use of DEM Data: A New Method. IEEE Transactions on Geoscience and Remote

Sensing, 52(1): 313-328

Zheng, C., and Wang, P. P., 1999. MT3DMS: A modular three-dimensional multispecies

transport model for simulation of advection, dispersion, and chemical reactions of

contaminants in groundwater systems; documentation and user's guide. Alabama

97

University. Technical report, U.S. Army Engineer Research and Development Center

Contract Report SERDP-99-1, Vicksburg, MS.

Zhu, X., 1982. Karst geomorphology and caves in Guilin, Annals of the Academy of

Geologic Science of China. Geologic Printing House, Beijing, 149-150

Zhu, X., 1988. Guilin karst. Shanghai: Scientific & Technical Publishers, 188pp.

Zhu, X., Wang, X., Zhu, D., and Qin, H., 1988. Study on Karst geomorphology and caves

in Guilin, Geological Publishing House, Beijing, 249 pp

Zhu, X., Zhu, D., and Ru, J., 1980. Geomorphology of tower karst and its development in

Guilin: Collected Papers of the International Geological Journal (5). Geologic Printing

House, Beijing, 98-108.

Zhu, X., Zhu, D., Zhang, Y., Lynch E.M., 2013. Tower karst and cone karst. In: Treatise

on Geomorphology, ed. Shroder, J.F. Academic Press, San Diego, pp. 327–340.

98

CURRICULUM VITAE

Wei Huang

Place of birth: Guiyang, the People’s Republic of China

Education

B.E., Hefei University of Technology, China, July 2002

Major: Biology Engineering and Food Science

M.S., Ibaraki University, Mar 2009

Major: Bio-Resource Science

Ph.D., University of Wisconsin-Milwaukee, Dec 2014

Major: Geography

Dissertation Title:

Spatial Dimensions of Tower and Cockpit Karst: A Case Study

of Guilin, China

spatial dimensions of tower karst and cockpit karst: a

Documents